Archive file of messages sent to genweb@ucsd.edu during November 1994.



          Wednesday, November 2, 1994 2:06:58 AM
          GenWeb Item
  From:           Herbert Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        Naming
  To:             GenWeb
Brian is discussing the problem how to make cross-pointers into several
data bases provided by data providers. This is the central question we have
to solve.

I prefer a non-centralized approach. Each provider should have a local
URL-scheme. We altogether should make a proposal how the typical scheme should
look like.
The arguments make it obvious, that the name should be a central part of the
genalogical URL. If we have several people of one name in a data base, we
have to find proposals for providers. A first proposal, which does not work
generally, is, to add birth data. This may be corrupted, if even several
people with same name and same birthday exist. This is open, if we don't know 
the birthday. 
The latter is the easy problem: give a count number (which has to be stable!).
The former could be solved likewise: give an additional count number.

I believe, that the decentralization solves the problems for us: It is not very
probable, that we have lots of same-named people in one database with exactly
the same birthdate. It occurs, that we have people in one data base with same
name and unknown birthdate and we know, they lived in around the same time.
Some times there are difficulties to separate them even. A local counter is a
solution (the only one?)

My proposal is: 
http://host/cgi-bin/genweb/T=/N=/F=/B=/C=/?LP
all parts between `genweb' and `?LP' are optional. In extremum, C=index will
serve.

T is for title
N is for family name 
F is for first name
B is for birth
C is for the local counter/index

Which char is good for `unknown'? `_'? `@'? `<>'?

Do we need a field for birth place? for fathers name?
Do not forget that we do not propose a global identification of people but only
an index into a data base.



          Wednesday, November 2, 1994 2:23:24 AM
          GenWeb Item
  From:           Herbert Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        job of ged2html
  To:             GenWeb
Translation in different languages.
If this will not be the job of ged2html, who else should do that? The current
ged2html can be moved out by a report program, as I have shown.
The report program can even deliver various name forms and presentations.

What report programs cannot do, I'm not clear with. I believe, you cannot
couple 2 lifeline data bases:

Tom, can I do the following:

Find a person, find a NOTE-field which says: info is in data base X,
calling lines with this info, producing a result, importing the results into
the first incaranation of lines?



          Wednesday, November 2, 1994 2:38:00 AM
          GenWeb Item
  From:           Birger A. Wathne,Birger.Wathne@vest.sdata.no,Internet
  Subject:        Re: Naming
  To:             GenWeb

But if we have local counters, then we dont need the other ones....

I think we can leave the indexing scheme to each site, as long
as the site doesn't change their scheme.

The problem lies with the host part. References should be stable
even if the base moves to a new location. If we can get around that
problem, either by URN's (work seems to be in progress somewhere,
I hope to get some info soon) or by some centralized mapping
from database ID to current server host, the rest of the lookup
can be completely up to each provider. I may want to use a stable
reference number on every person in my base. You may want to index
on names, etc.

In Iceland, the first name is used for indexing, as any person's
last name depends on his father's first name. People in norway often
changed their last names depending on where they lived. I have two
birth dates, one correct (from birth certificates, etc) and one
official (used in social security number, etc). So there are lots of
problems that keep me away from using this info. Changing any of these
(name, birth date, etc) will change the locator for the person!!!!!
So I propose that each base has its own numbering scheme (based
on index numbers within the data, not the GEDCOM index numbers),
and that lookups are based on these. These index numbers should
be stable even if the data provider decides that the name is misspelled
or the birth date is incorrect.

As I have stated, I think that the problem with global location of a
person lies in the hostname used in the URL.  A locator to a given
person should not be bound to the host the base resides on. The base
could move. The person could get moved to another base, for that
matter, but that problem opens the same can of worms on a new level. If
we want to be able to follow a database when the complete base moves
(If I move to a new job...), we need a global locator for the database.
But if we want to follow a single person if responsibility for this
person is moved from one provider to another, we need global locators
for each database entry. These locators should still not be bound to
the data on the person. The only solution I can see for this, is to
keep tabs on each person in some central (but perhaps still distributed
and synchronized) archive. If we want to sole the problem of truly
global locators on each person then I think we have a huge job ahead.
We may just want to see what the WWW community is doing to solve this
problem on a more general level.

In the meantime, I guess each site could develop local indexing mechanisms,
and be prepared to change links to external databases when something
more stable is found. Select your own indexing schemes with care.

So my recommendation would be to use something like
http://host/search-program/INDEX=MY1234?Lookup

Make shure the index number will remain stable....



Birger



          Wednesday, November 2, 1994 2:56:18 AM
          GenWeb Item
  From:           Brian Tompsett,B.C.Tompsett@computer-science.hull.ac.uk,Internet
  Subject:        Re: Naming
  To:             GenWeb
 Herbert Stoyen ommitted to reference my draft discussion document on the
matter:

http://www.dcs.hull.ac.uk/public/genealogy/DistributedGenealogy.html

 [I hadn't announced it yet as I still have a huge amount of discussion to
  append to the paper.]

 Brian



          Wednesday, November 2, 1994 7:29:13 AM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Re: Naming
  To:             GenWeb
Birger Wathne writes:
>So I propose that each base has its own numbering scheme (based
>on index numbers within the data, not the GEDCOM index numbers),
>and that lookups are based on these. These index numbers should
>be stable even if the data provider decides that the name is misspelled
>or the birth date is incorrect.

I agree with Birger that database index numbers are probably more
stable than names (which are subject to ambiguities as well as
changes and arbitrary spellings), and therefore suitable in URLs.
However, no database maintainer should be bound by this conclusion,
and client software should not depend on GenWeb URLs having a
particular syntax.

In order to find the URL or index number for a person in a
particular database, that database should be accompanied by
an index where individuals can be looked up by full or partial
names (as well as other search criteria).

>So my recommendation would be to use something like
>http://host/search-program/INDEX=MY1234?Lookup

I'm confused by this choice of syntax.  While the HTTP protocol
doesn't enforce any particular interpretation of the various fields,
a common convention is that the part found to the left of the "?"
refers to a preexisting "document" (whether physical or logical)
with the "/" indicating hierarchical subdivision of some kind,
and the part found to the right of the "?" is an arbitrary search
string, which may retrieve whatever document the server finds
most appropriate for the moment.  Therefore, if I had made the
choice, I would have written it something like

	http://some.dom.ain/ll-gw/lookupindex?MY1234

I don't consider this important enough to suggest changing
existing implementations; I leave it to those doing the work
as mere food for thought.  It will of course work either way.

>As I have stated, I think that the problem with global location of a
>person lies in the hostname used in the URL.  A locator to a given
>person should not be bound to the host the base resides on. The base
>could move. The person could get moved to another base, for that
>matter, but that problem opens the same can of worms on a new level. If
>we want to be able to follow a database when the complete base moves
>(If I move to a new job...), we need a global locator for the database.

For the domain address part, maybe a directory of DNS aliases would
solve the problem (in the interrim, until the URN or a corresponding
solution is fully established)?  This requires at least two domain
server maintainers to offer service for a new domain, say GENWEB.ORG
(for example).  This domain would contain a name for each database
defined as an alias (CNAME in DNS lingo) for a physical host.  As
the database maintainer moves, the alias name can be redefined and
all proper URLs will remain as valid as they were before.  For instance,

	http://wathne.genweb.org/cgi-bin/ll-gw/BASE=wathne/...

would refer to your existing database, by means of "wathne.genweb.org"
being an alias for "sdvest.vest.sdata.no".  In the same fasion,

	http://royal92.genweb.org/cgi-bin/ll-gw/BASE=royal92/...

with "royal92.genweb.org" being an alias for the same host would
be the standard reference to your other database.

Note that the database name is present twice in the URL.  This
redundancy is intentional, as the HTTP server cannot know by what
alias domain name it was identified (the HTTP protocol provides it
with the "/cgi-bin/ll-gw/BASE=..." string only).  This enables
Birger to have only one of his databases (say, royal92) moved to
another server.  One alias would be changed to indicate the new
location, and the other would remain unchanged.  Changing the
alias would be the job of the GENWEB.ORG domain administrator.

Rather than having a completely new domain registered with the NIC,
a simple solution would be to "borrow" part of an existing domain
for this purpose.  Are there any other DNS server maintainers out
there concerned with the GenWeb?  Could we have suggestions for an
existing domain which might offer support for GenWeb in a subdomain?
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE



          Wednesday, November 2, 1994 4:41:25 PM
          GenWeb Item
  From:           Bill Harten,wsharten@comnet.com,Internet
  Subject:        External Links
  To:             GenWeb

Please let me introduce myself.  I am the product manager currently 
responsible for GEDCOM at the LDS Family History Department, and I am the 
originator of the GEDCOM technology, as well as the GEDCOM-oriented database 
technology currently used in our large CD-ROM FamilySearch databases.

I have been watching GenWeb from the beginning with intense interest, and 
believe the basic concept to be crucial to the future of genealogy.

The progress made by the GenWeb discussion group is most impressive.  It is 
very encouraging to see such positive dialog and to watch problems being 
solved by effective collaboration.

I hope my time will permit me to participate with you as you work through 
the numerous issues to be faced.

As a starter, let me provide some information on a point I just read from 
Herbert Stoyan:

>Additionally, gedcom lacks the following things:
>...
>external-db-links (for people which are mentioned in one db but the full 
>info>is in another db)
>
>

External references are critical.  There has to be some way to refer to a 
record in another database by means of a _permanent_ record identifier, 
combined with a permanent resource identifier, which could be thought of as 
a kind of ISBN (don't know the acronym) book identifier.  The 
resources/records thus identified have to be able to move around as systems 
come and go, without invalidating pointers in other databases/pages.

We anticipated this requirement years ago, and reserved a special symbol to 
support it in The GEDCOM standard, draft version 5.3.  On page 11, in the 
definition of cross reference identifiers (opt_xref_id) it says "...The use 
of the colon(:) character is reserved."  It is reserved precisely for the 
purpose described in the preceding paragraph.  The info on the left side of 
the colon is for a resource identifier.  A record number, unique and 
permanent within the database resource, goes on the right.  Earlier drafts 
explained this in more detail, but I removed the detail to avoid confusing 
anyone.

From my understanding of the web's URL format specifications, I now see that 
the pound sign(#) could be used instead of the colon--URL's had not been 
defined at the time this concept became part of our architecture.

We started assigning resource identifiers and permanent record numbers to 
our large databases several years ago, seven of them so far.  I am sure we 
need to have further discussions of the best way to integrate this 
requirement in terms of existing, planned, or needed resource registration 
mechanisms.  This is not an announcement that we plan to put our databases 
on the web in the near future, but we plan to be prepared to do so if that 
time comes.

I would appreciate ideas on how this should work.

Best regards, and congratulations on your progress.

Bill Harten
Manager of Architecture, LDS Family History Department
wsharten@comnet.com
phone 801-240-5225



          Thursday, November 3, 1994 2:02:28 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        external pointers
  To:             GenWeb
Nice, that Bill Harten wrote to us.
 Lets come to the point:

Can I have a family definition

0 @Fxxx@ FAM
1 @Ixxx@ HUSB
1 @#http://host/cgi-bin/ll/B=base/I=index/?LookupInternal@ WIFE

??

Bill did note tell about selectors for predecessors/postdescessors in office
Are there:

1 PRE @Ixxx@

or

1 POST @Ixxx@

(what else is free for that purpose?)??

A last question:

What can be used as individual identifier:
Only `@Idddd@' (d=digits)? Can I use: @Ihabsurg222333@
??



          Thursday, November 3, 1994 6:21:33 AM
          GenWeb Item
  From:           Tim Doyle,tdoyle@netcom.com,Internet
  Subject:        Setup question
  To:             GenWeb
I hope I'm not out of place asking a newbie question like this here.  I 
would very much like to setup my GEDCOM information on the web.  I use 
Netcom, and as I understand it, this means that users must ftp to my data 
instead of http.

I am attempting to install the ged2cgi generated files, but have come up 
against a problem and would like to know if what I am attempting is even 
possible.  Since the ged2cgi stuff uses perl scripts to access the index 
files, can users run these perl scripts even though they are ftp-ing to 
my site? How would the perl script run the perl program listed on the 
first line of the script?  Do the scripts need to be 'compiled' in some 
way?  Am I totally lost, or has someone else come up against this problem 
before?

Any help would be greatly appreciated.

By the way: the problem that I am getting now is when I click on any of 
the links to the perl scripts.  I get 'Unable to locate file'.

                                             tdoyle@netcom.com




          Thursday, November 3, 1994 7:48:31 AM
          GenWeb Item
  From:           Gene Stark,starkhome!gene@sbstark.cs.sunysb.edu,Internet
  Subject:        Naming
  To:             GenWeb
After reading a number of the recent messages on this list regarding
naming in GenWeb, I have begun to think that a model similar to that used
by the Internet Domain Name Service (DNS) might be the most reasonable.
This idea was suggested by Anders Andersson  .
Actually, he suggested the use of the DNS service itself as an interim
solution until a better one is invented, but I doubt if the current DNS
software can be "bent" far enough to be useful in GenWeb, as it was designed
with a somewhat different application in mind.

The basic idea, which is the same as the DNS, is that there will be a
bunch of servers running on the net, with the general responsibility
of providing lists of URLs in response to queries.  They could be
hierarchically organized, as in the DNS, with one or more "root" servers
that have knowledge of how to access servers at the next level down.
Queries issued to one server would propagate toward the root until a
server with "authoritative" information is located.  (One potentially
difficult issue here is that, in contrast to the Internet, there may be
multiple "sources of authority" for a given resource.)

More specifically, the responsibilities of the servers would be as follows:

1.  Translation of global "person identifiers" into URL's.
	The global identifiers would be stable, and therefore would have
	to be assigned by a central authority.  There could be more than
	one series of global identifier, but there probably shouldn't be
	too many (i.e. thousands) of them.  The identifiers, such as the
	"Ancestral File Numbers (AFNs)" assigned by the LDS to their
	"Ancestral File" database, are an example.  I believe there is
	a separate numbering scheme for the IGI, etc.
	These global identifiers would serve a purpose similar to that
	served by domain names in the internet: a stable identifier that
	can always be used to locate a previously identified resource when
	required.

2.  Lookup, based on personal data (name, birthdate, etc.), of global
	identifiers and URL's.  Matches need not be exact, so there would
	have to be some way to limit the number of matches returned in
	response to any given query, and in case of too many matches,
	to return information the client could use to further restrict
	the query.  Responses to such a query would be a list of URLs
	and global identifiers (if assigned), together with the personal
	data on which the match was based.

3.  Relaying queries to another server, if authoritative information
	cannot be found.  Queries would by default propagate toward
	the root, but could propagate toward the leaves in case indexing
	information is encounted that suggests that this might be
	helpful.

The idea in this scheme is that anyone who wants to can run a server
and prime it with translation information about their own local database.
For this information to be accessible from elsewhere, it will be
necessary for the server to be registered with a server closer to the root.
Indexing information will have to propagate regularly toward the root in a
regular fashion, so that queries arriving near the root can be directed back
toward as few leaf nodes as possible.  Servers would cache this indexing
information, and as in the DNS it would have automatic expiration (but
probably with much longer expiration times, e.g. weeks rather than hours)
to prevent an accumlation of outdated information.

An important aspect of this scheme is the separation of the name service
from the actual Web data.  The actual authoritative information about the
URL corresponding to a particular resource would be known only to the
individual servers near where the data is stored.  This would permit an
individual GenWeb site to make changes to the actual location of their data
without having to constantly reregister with a central site.
Web documents would still contain URL's but  but failure to contact a URL
would trigger a query to a server to determine if the data has moved.

There are plenty of issues that are problematic in the above proposal,
but maybe it could be a starting point.  The main problem I see right now:
How to use indexing information to effectively filter queries so that
they propagate only toward the servers that are likely to actually
have the requested data.  One thing I forsee here is that there
is likely to be only a few "archives" of data on any given individual,
but many more "mirrors", which are people who have copied this
data into their own local database.  There should be a way to
distinguish between "archives" and "mirrors", and have a query
propagate only toward the archive servers for that individual.
Also, maybe there should be a way that access to a mirror
automatically triggers a check to see if the archive copy has been
updated.

							- Gene Stark




          Thursday, November 3, 1994 7:51:15 AM
          GenWeb Item
  From:           Scott McGee,smcgee@microware.com,Internet
  Subject:        information request
  To:             GenWeb
I just heard that this is a genweb discussion group. I would like to join
if possible. I am currently attempting some limited genweb type stuff myself
based on some of the genweb ideas I've seen on the net. My site does not yet
permit outside access so I can not export my experiments, but would like to
know where things are going and keep up as best I can. 

If someone could tell me how I might join the discussion group (mail list?)
I would be gratefull.

Scott 

        Buttered bread always lands   *   Would YOU mistake these
        butter side down! (Unless     *   opinions as anyone`s
        it sticks to the ceiling!)    *   but my own?
                  smcgee@microware.com (Scott McGee)



          Thursday, November 3, 1994 9:57:54 AM
          GenWeb Item
  From:           Birger A. Wathne,Birger.Wathne@vest.sdata.no,Internet
  Subject:        Re: Naming
  To:             GenWeb

I really like your proposal.

I am not shure I would be allowed to use our domain this way, as we are
a commercial organization, and security implications would have to be
checked. I have enough problems keeping the httpd running :)

I may have to make a copy of my bases anyway (And put them on a host
outside our firewall), but I still think running straight from a
database will give a lot of benefits. I'm now looking into letting
users change parts of the layout (waiting for a ged2html with support
for this). Simple things like selecting tag language, selecting date
formats, and choosing if they want lastnames uppercased. But if all
versions should be saved as separate HTML files........

I just found the latest URN draft at .
I'll have to read it carefully, and then do some thinking. Perhaps on
the plane tomorrow? A plane has just been hijacked here in norway, so I
guess my plane will be safe :)

From a quick reading, I think we may just go ahead with your scheme.
Perhaps some organization with interest in genealogy could offer
a genweb subdomain? We could build WWW forms and mail responders to
register new bases into the DNS automagically. It would even fix the problem
of overhead by going through a central host. The DNS lookup will be cached
locally by the resolver.

My reason for putting the variable part before the '?' was that I seem
to be able to get the part up to the '?' without URL interpretation
being done (interpretation of '%' sequences, etc), so I don't have to
check for special characters when generating the URL's. If someone
creates a database with some reserved character in the database name,
etc....

It's nice to see that collective thinking can work.
But now, I'm off for dinner.


Birger



          Thursday, November 3, 1994 2:17:53 PM
          GenWeb Item
  From:           lees@dg-rtp.dg.com,Internet
  Subject:        What is the objective of genweb?
  To:             GenWeb
Bill Harten of the LDS Family History Department (who just introduced
himself on this mail list yesterday) pointed me to the genweb list
several weeks ago.  I have been reading genweb now since early October.

My question is: what is the objective of genweb?  Is there any written
goal?

From reading the messages it looks like it is to create a WWW based
distributed data base of individuals, their family relations to each
other and vital statistics and other information about those
individuals.  Is this accurate?  Is the data base to include only the
*results* of research, or is it to eventually include source material
such as indexes, abstracts and transcriptions of documents?

Having a data base of individuals is great for sharing information
between researchers working on a common line, and in itself a *very*
worthy goal, but online source information would allow me to break
*new* ground more effectively.

A self introduction:

I work for Data General as a Principal Software Engineer developing
and maintaining the X Window System on our Unix workstations.  In
my spare time I have been developing a genealogy data base that is
structured around GEDCOM (and has many of the same features as
LifeLines).  I have also been working with Bill Harten and the LDS
Family History Department on revising the GEDCOM standard.

-----
Scott E. Lee        (919)248-6370      lees@dg-rtp.dg.com
Data General Corp.  (919)248-6108 FAX  {world}!mcnc!dg-rtp.dg.com!lees
  "It's hard to tie your shoes when a cat is watching!"  -- Me



          Thursday, November 3, 1994 2:44:16 PM
          GenWeb Item
  From:           Bill Harten,wsharten@comnet.com,Internet
  Subject:        Re:  External Links
  To:             GenWeb
Anders Andersson wrote me privately, Jim Eggert said the same, and I pass it 
on as instructive:
>
>>From my understanding of the web's URL format specifications, I now see that 
>>the pound sign(#) could be used instead of the colon--URL's had not been 
>>defined at the time this concept became part of our architecture.
>
>I don't think this is a good idea.  In a URL, the pound sign indicates
>a label in a document where the browser should position its point of
>view.  Note that the #-part of the URL is never passed from the client
>to the server as part of a request.  Instead, the client requests the
>entire document (the URL without the #-part) and finds the proper
>location in it all by itself.
>
>If you are making an analogy between a database resource and the part
>of the URL to the left of the pound sign, the client software would
>have to retrieve the entire database simply to access one entry.
>This is of course not practical.
>
>If you really need a corresponding mechanism in the URL, I believe the
>question mark (?) most accurately describes what you want for a reference
>in GEDCOM.  To the left of the "?" is a document (it could be a large
>database or a single page of text), and to the right is an arbitrary key
>or combinations of keys for a database search or lookup.  In contrast to
>the #-label, this lookup is done in the server end.
>
>However, I'm not sure of what you are trying to do, and I don't think
>GEDCOM should be too much dictated by the intrinsics of URL syntax.
>In a URL, a colon (:) may appear in at least two different places:
>
>	http://www.foo.bar:8000/database?index#label
>
>While the first colon is mandatory and separates the access method
>("http") from the rest of the URL, the second colon is optional and
>separates the server name from the port number.  Obviously, a URL
>parser is able to distinguish between these two interpretations of
>a colon, due to the rest of the syntax.
>
>I have the impression that you have planned to use a syntax like
>"database:123456" for a reference to an external database, right?
>Do you plan to impose some restrictions on the syntax of either the
>database field or the index field?  Should the database resource
>identifier conform to a centrally managed registry of resource types,
>or should it merely be a string of ASCII characters to be interpreted
>by the user's software?  Should the index be numeric only, or could it
>be an arbitrary alphanumeric string?
>
>I mention these issues simply to suggest that you will need some way to
>translate between your identifier strings and corresponding strings in
>the outside world.  For instance, if one database is given the pretty
>unintuitive name of "Catch#22", you must rewrite it if you are to use "#"
>as your field separator.  The same goes for colon in the case of a URL as
>the database identifier.  Then you have the possibilities of blank space
>etc.  You could enclose the database ID between quotation marks:
>
>	XREF "http://www.foo.bar:8000/database":index
>
>However, then you need to take precautions against future database having
>quotation marks as part of their names...  It's always possible, although
>you may end up with an inconvenient syntax.
>
>Having written this, I think I better understand what you wrote, and
>using the pound sign as your separator might not be too awkward after
>all.  Just make sure not to confuse it with the URL pound sign, which
>has a different meaning.  Fortunately, it's unlikely to appear as part
>of a database URL.  OK, what kind of common misunderstandings will
>
>	XREF http://www.foo.bar:8000/database#index
>
>lead to among not-too-experienced computer genealogists?  It looks
>very much like a URL, but it isn't, and it should probably be
>translated to "http://www.foo.bar:8000/database?index" or something
>similar before actually being used on the Web.  The British experience
>with reversed domain addressing comes to mind (ask me if you don't
>know what I'm referring to)...
>
>Regards,
>Anders Andersson, Dept. of Computer Systems, Uppsala University
>Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
>Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE

Anders, I think you understood my intent correctly, and that perhaps I 
misunderstood the difference between the # and the ? in URL syntax.  I 
thought the symbols after the # were passed to the server, to be resolved 
according to a meaning known to the server.  My intent is to use whatever 
the URL syntax would specify to identify a record within a database.

I do not understand the web protocols well enough yet to say exactly how 
URLs might be used correctly in GEDCOM or in a web-oriented database.  This 
is important.  Your comments here are helpful.  I would appreciate more 
information about the British work you mentioned, plus anything else that 
would help me/us more quickly understand these issues.

Best regards,

Bill



          Thursday, November 3, 1994 11:53:25 PM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        Net pointers
  To:             GenWeb
If I understand WSHarten right, we could use URLs inside record identifiers.

i.e.: a person could have family pointers across the net:

1 FAMS @http://....@

a family could have person pointers across the net

1 HUSB @http://....@

This would work, if there are no @ inside URLs -- are they? Is there some 
character which is not permitted inside URLs?



          Friday, November 4, 1994 3:09:05 AM
          GenWeb Item
  From:           Brian
Tompsett,B.C.Tompsett@computer-science.hull.ac.uk,Internet
  Subject:        Re: External Links
  To:             GenWeb
 It seems to me that we are iterating around an answer.

 If we combine the concept of a genweb domain, GEDCOM syntax and discussions
on record ID we can solve these together. Can I suggest the following:

  XREF:LINK:=
    :

   Here we use the same syntax implied in GEDCOM5.3 (page 45).

   Database:=
     A name of a GenWeb database registered in the GenWeb DNS domain
     to retrieve records using the http protocol. This will access
     the data by the URL
       http://Database.genweb.org/cgi-bin/genweb/Database?

   !ID:=
     A record ID as supported by the specific database being accessed.
     This may be an integer or a character string. If a string it may
     not contain the characters ":" (and some others...)


     It is required that the record ID will always retrieve the same
     data record irrespective of future changes to the database.

  *Now* I see one *Big* problem. In the GEDCOM spec the size of XREF's
  is limited to {Size=1:15} which does seem to constrain the type of names
  you can use for the Database and record ID! :-)

  If I finish writing my discussion document to include details on how you
  can implement things under the various database schemes to have unchanging
  record ID, as well as suggesting desired minimal searching functions that
  can be used to aquire a record ID from any database, isn't that everything?

  The only problem now is that a database can only be supported at a site
  where the data provider is of Web Manager status or above. I would have
  trouble getting myself in a Genweb domain (political reasons) but I'd
  be prepared to stand up and be counted on the grounds of academic research.
  Others (who have really good data to contribute) are not so fortunate.
  Many oly have ftp, or not even that. This means the scheme wont work for them
  as it makes assumptions about protocol and netstatus.

  The converse argument is; do we really want REFERENCE sources of data at
  sites where genweb does not have the support of those in authority. The
  concept of a long-lived ID is a bit suspect. (e.g. Sorry I was over quota
   and I had to remove a few things.....)

  One things I think has not been aired yet, is the fact we will need a
 "see also" link. With a distributed system we cannot guarantee a single
 record on an individual. Say everyone has Queen Victoria or Charlemagne
 in their database ('cos its fun to be desended from 'em) then people will
 have to point their Vicky record at the others,

  Brian


          Friday, November 4, 1994 4:18:20 AM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Re:  Net pointers
  To:             GenWeb
Herbert Stoyan writes:
>1 HUSB @http://....@
>
>This would work, if there are no @ inside URLs -- are they? Is there some 
>character which is not permitted inside URLs?

"@" characters are considered "safe" and are allowed inside URLs, for
instance in the "path" and "search" parts.  In addition to that, they
have a special interpretation within FTP, MAILTO, NEWS and TELNET URLs
when they appear before a domain name.

In short, the characters A-Z, a-z, 0-9, "$", "-", "_", "@", ".", "&",
"+", "-", "!", "*", '"', "'", "(", ")", and "," may appear anywhere in
the path or search segments.  The characters "=", ";", "/", "#", "?",
":", and " " (space) are reserved and may only be used where specifically
allowed by the standard, if at all (the space is never part of the URL,
but may surround it).  The "%" character may be used as an escape prefix
followed by two hex digits to encode other characters.  The characters
"<", ">", "[", "\", "]", "^", "{", "|", "}", and "~" are disallowed
entirely, as are any non-printable (00-1F, 7F) or non-ASCII (80-FF)
characters.

(Note that according to the specification, the "=", which is currently
in use by some of the GenWeb test gateways, is actually reserved for
use in FTP and WAIS URLs.  However, few implementations follow the
specification to the letter, and therefore a lot of "disallowed"
characters are appearantly permitted.  The specification doesn't
require applications to return an error message in each such case,
but on the other hand such usage is not supported, and may become
inconvenient with future development of the standard.)

While the "%" prefix is intended primarily to encode characters which
are disallowed in URLs (such as "%09" for the TAB in Gopher strings),
it may also be used to encode plain normal characters (an example is
"%40" for "@") if for some reason the environment where the URL appears
has a problem with that character.  In theory, it seems this would apply
well to the GEDCOM context.  However, I just made some experiments with
this, and it seems that at least Mosaic has problems interpreting "%40"
as a regular "@" other than in Gopher menus (I don't know why).  Further
study of the issue is therefore required.

Note also that the reserved characters listed above may not be encoded
in this fashion.  See the 
URL specification for a more in-depth description of URL syntax.

When rereading this document, I notice that the "#" isn't formally part
of the URL, but of the URI (Uniform Resource Identifier).  Thus, a URI
may consist of a URL followed by "#" and a fragment ID (label).  In the
future, a URI may also consist of a URN (Uniform Resource Name/Number).
Maybe we should state that the item of interest for GEDCOM is the URI
and not just the URL?
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE



          Friday, November 4, 1994 8:01:46 AM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Re:  Naming
  To:             GenWeb
Gene Stark writes:
>After reading a number of the recent messages on this list regarding
>naming in GenWeb, I have begun to think that a model similar to that used
>by the Internet Domain Name Service (DNS) might be the most reasonable.
>This idea was suggested by Anders Andersson  .
>Actually, he suggested the use of the DNS service itself as an interim
>solution until a better one is invented, but I doubt if the current DNS
>software can be "bent" far enough to be useful in GenWeb, as it was designed
>with a somewhat different application in mind.

I'd like to point out that I don't intend to tweak DNS beyond its
present application, and that it in no way solves all our problems.

While the CNAME capability in the original DNS specification may have
been designed with host name changes and transitional names in mind,
it is used today to provide machine independency for services such as
FTP and HTTP.  Many (I'd believe most) names of the WWW.site.domain
kind are aliases for some physical host.  In case the site manager
decides to move the HTTP server to another machine, he can simply
redefine the alias to identify the new server, and every existing URL
will remain as valid as they were before the move.

There is no requirement that the service alias and the name of the
machine belong to the same parent domain.  If Acme, Inc provides
Frobozz Ltd with network services including an HTTP server,
WWW.Frobozz.Com may have a CNAME of somehost.Acme.Com.  In case
Frobozz Ltd decides to dump their service provider for another one,
they can simply have their CNAME changed.

I'm suggesting using the same mechanism for the GenWeb, simply to
avoid having to rely on the names of non-affiliated organizations.

This solution takes only the domain part of the URL into account.
In order to provide stability for the rest, i.e. access method and
path name, other solutions (such as cooperation from the server
manager) are required.  However, the DNS CNAME method gives you
great freedom in the choice of server manager, and even if you
have to maintain your database on a particular server where you
have little influence over the naming structure, I imagine that
you could arrange with a cooperative GenWeb server to provide
arbitrary URL forwarding.

For an example of how this might work, please have a look at my page
GenWeb server
canonical domain names and try the links (you should discover at
least one potential problem which I haven't addressed yet).

However, you are quite right that playing around with the DNS is
not a good long-term solution.  We need resource identifiers that
are reliable, flexible, and independent of genealogically irrelevant
things like computer names and network infrastructure.  I can't say
that I read and understood your suggestions thoroughly, but I think
they may constitute the next step in the process, after this one.
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE



          Friday, November 4, 1994 8:09:27 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        goals
  To:             GenWeb


There has been a mail of G.Hofman which was based on a paper of him in COMPU.GEN
I add this to this mail.

Some more goal defining mails have been sended. May, we have to work on a common
document.

Presenting the results is definitely only one of the goals. A part of this
presentation goal is the distributed data base of individuals (vital statistics,
family relation, biography, pictures, etc.). Therefore, related source info
is important.

We discussed already possibilities to collect individual related research
attempts of genealogists.

It is clear, to induce the motivation for having the other tools, a common
presentation tool would be great.

[enclosure omitted here]



          Friday, November 4, 1994 8:13:13 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        Net pointers
  To:             GenWeb
Ok Brian, the proposal is good. I do not understand, where the XREF can be
positioned. After the INDI line?
What does it mean, Tom, that you don't want `to maintain' cross-reference keys?
Don't you permit cross-reference links at all? Don't you want it to be coded
inside the keys. Or what?



          Friday, November 4, 1994 8:24:03 AM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Re: External Links
  To:             GenWeb
Brian Tompsett writes:
>       http://Database.genweb.org/cgi-bin/genweb/Database?

>  Many oly have ftp, or not even that. This means the scheme wont work for them
>  as it makes assumptions about protocol and netstatus.

This scheme won't enable us to use the same URL to access a database
via FTP one day and via HTTP the next, that is correct.  Just as the
"/cgi-bin/genweb/Database/" part has to stay the same for a given
database, the access method must remain the same.  However, it need
not be HTTP.  It could just as well be FTP or Gopher, as domain names
are used in those URLs also.  This is of course not a very pretty
solution, but something we will have to live with as long as we use
URLs (Uniform Resource *Locators*).
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE



          Friday, November 4, 1994 8:41:22 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        Expansion of Demo
  To:             GenWeb
I have installed a first engine for direct person access, a time-oriented
access, and a name-oriented access (gives lists of people with some name).

The problem is, its running under lines2.3.6, not lines301.

Look at http://faui80.informatik.uni-erlangen.de/ww-person.html



          Friday, November 4, 1994 8:57:59 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        Problems with Zombie processes
  To:             GenWeb
We had some www-visitors here and they looked into my system. Some of them
moved away and killed their application. lifelines processes were still up
and converted into defunct-processes. How do you handle the problem?
Garbage collection of processes?


          Friday, November 4, 1994 7:50:27 PM
          GenWeb Item
  From:           fr@icdi10.compu.com,Internet
  Subject:        while we're on the subject
  To:             GenWeb

 SGML Open Aims at New Target: the Web
 -------------------------------------
      Committee will try to facilitate exchange of documents

 New support for facilitating the exchange of documents between
 the World-Wide Web (WWW) and corporate information databases
 is coming from SGML Open, a consortium of vendors whose products
 support the Standard Generalized Markup Language (SGML).

 SGML is the international standard for the electronic representation
 and interchange of document-oriented information among users,
 platforms, and applications. Published by the International Organization
 for Standardization as ISO 8879, SGML is overseen by an international
 working group but is not owned by any organization.

 SGML Open announced at the recent WWW Conference in Chicago the
 formation of a committee on SGML and the Internet. Its objective
 is to provide technical resources and support for the exchange
 of WWW documents that employ the Hypertext Markup Language (HTML),
 an application and subset of SGML.

 Version 2.0 of HTML is now being finalized by the Internet Engineering
 Task Force (IETF). That version is intended to stabilize the
 relationship between HTML and SGML by providing a document type
 definition (DTD). The DTD, a key part of SGML, specifies tags
 that are legal within an application and the relationship of
 the data elements to each other. The previous version of HTML
 did not have a DTD, and therefore was not "real SGML" according
 to Mary Laplante, SGML Open executive director.

 In order for corporations to be able to exchange documents and
 make them available and interchangeable between their databases
 and the Web, the documents need to conform to SGML and the SGML
 and HTML standards need to be aligned. SGML Open is promoting
 the widespread use of SGML on behalf of its members, who make
 and support SGML products and services.

 "We certainly want to take advantage of the interest in the Internet,"
 Laplante says. "The reason we formed the SGML and the Internet
 Committee was to try to work with WWW users to assure the use
 of SGML directly on the Internet. That will be one of our primary
 goals for the next year." One company, SoftQuad, of Toronto,
 Canada, announced a product at the Web conference that facilitates
 the use of SGML documents on the Web.

 "If all you want to do is publish on the Web, what you've got
 today is fine, but corporations usually want to do other things
 with their data," Laplante says. "They want to publish in on
 paper, put it on CD-ROM, store it on a database, use it in real
 time, internally. HTML is just not sufficient for that. So by
 making HTML documents available in real SGML tools, you really
 extend the utility of those documents, and their flexibility
 and accessibility and all the other wonderful things that SGML
 theoretically provides for you."

 Chaired by Steven J. DeRose, chief scientist at Providence, RI-based
 Electronic Book Technologies, Inc., the committee's agenda will
 also include:

      o  Helping organizations extend the utility of their HTML documents
 through the use of SGML tool sets.
      o  Working with WWW users to bring support for longer documents
 and greater graphic presentation capabilities to the Internet.
      o  Offering SGML Open's technical resources for review of the
 future HTML 3.0 in collaboration with the HTML working group
 of the IETF.

 Formed in 1993, SGML Open focuses on promoting product interoperability
 through technical resolutions, educational activities and marketing
 studies. One of the largest obstacles to universal adoption of
 SGML is that it's expensive to implement, especially to convert
 legacy documents to SGML, according to Laplante. "It's still
 expensive, still complex, still requires expertise, and it's
 still hard," Laplante says. "By working together through a common
 practices approach, we're trying to make the products work together
 better, to put better user interfaces on them, and to drop the
 cost barrier to getting into SGML. Another issue, on the customer
 side, is the great cost of trying to deal with inconsistent data,
 which is what SGML was designed to address. Files can appear
 consistent from an appearance standpoint, but the electronic
 files are completely inconsistent, and that drives up the cost
 of conversion."

 *** For further information, contact:

      o  SGML Open, Coraopolis, PA; (412) 264-4258; laplante@sgmlopen.com
 (Mary Laplante); WWW server http://www.sgmlopen.org
      o  International Organization for Standardization (ISO), Geneva,
 Switzerland; +41 22 749 011
      o  The Internet Society, Reston, VA; (703) 648-9888;
 Email: isoc@nri.reston.va.us


 ------------------------------------------------------------

Fred



--
W. Fred Rump or CompuData, Inc.    | 'When we were young, we did not ask
26 Warren St |  10501 Drummond Rd  | questions; now that we're old, there's
Beverly NJ 08010 |  Phila. PA 19154| 		no one to answer them.'
609-386-6846 | 215-824-3000| fr@icdi10.compu.com or 73512.2234@compuserve.com



          Saturday, November 5, 1994 6:56:17 PM
          GenWeb Item
  From:           T.T.Wetmore,ttw@beltway.att.com,Internet
  Subject:        Re: External Links
  To:             GenWeb
I have not commented on this thread before.

I am confused about all the concern about what exactly can be placed in a
GEDCOM external, cross reference link.  The value in the link does not have
to be in the syntax of a URL value to be easily translated into one.  Nor
does the URL link have to appear in the link value itself; it could be kept
in another line, say one with the URL tag.

Tom Wetmore, ttw@beltway.att.com



          Sunday, November 6, 1994 9:46:24 AM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Re: External Links
  To:             GenWeb
Tom Wetmore writes:
>I am confused about all the concern about what exactly can be placed in a
>GEDCOM external, cross reference link.  The value in the link does not have
>to be in the syntax of a URL value to be easily translated into one.  Nor
>does the URL link have to appear in the link value itself; it could be kept
>in another line, say one with the URL tag.

Quite true.  What I was getting at, is that in order to include an
arbitrary URL in a GEDCOM field, all potential URLs must be handled
by the syntax chosen.  As for putting the URL in another record, that
only changes the precise syntax requirements, but it doesn't eliminate
them.

Due to the uncertain future of existing URLs, I'm tempted to suggest
keeping them out of the GEDCOM databases altogether, and establish an
external mapping between database identifiers and URLs.  Such a solution
would make a future conversion from URLs to URNs or any other resource
identification scheme much simpler.
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE


          Monday, November 7, 1994 2:03:09 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        external links
  To:             GenWeb
There are 2 issues: syntax of external pointers and management.

The syntax is still be open and may be a URN, or else.

The place where the external pointers are stored has to be the database.
What else is possible? If I use a internal walk through a database and
find a person, about which real info is outside the db, this has to be
stored in the db. How can a `external mapping between database identifiers
and URLs' make any sense (if not finding the real URL for some form of
external pointer)? 

Lifelines removes all original person identifications and counts from
1. If someone has person identificators (@I...@) with content, this is
lost. Why?



          Monday, November 7, 1994 12:28:37 PM
          GenWeb Item
  From:           lees@dg-rtp.dg.com,Internet
  Subject:        Re: External Links
  To:             GenWeb
>   One things I think has not been aired yet, is the fact we will need a
>  "see also" link. With a distributed system we cannot guarantee a single
>  record on an individual. Say everyone has Queen Victoria or Charlemagne
>  in their database ('cos its fun to be desended from 'em) then people will
>  have to point their Vicky record at the others,
> 
>   Brian

GEDCOM defines an ALIA (alias) tag that, according to the standard:
"... indicates that another record is suspected of being the same
person."  This seems to be just what you are looking for.  The
standard goes on to say that when you confirm the suspicion you
should merge the two records -- possibly good advice for all of
those Queen Victoria records.

Scott



          Monday, November 7, 1994 8:07:50 PM
          GenWeb Item
  From:           T.T.Wetmore,ttw@beltway.att.com,Internet
  Subject:        Re:  external links
  To:             GenWeb
>Lifelines removes all original person identifications and counts from
>1. If someone has person identificators (@I...@) with content, this is
>lost. Why?

The key values used as inter-record pointer belong to the database, not to
the user.  Programs are free to change those values as they see fit.  The
LL approach is to simply start at I1, F1, E1, S1, X1, for the five
different kinds of records, and count upwards as new records are added.

There are other ways to give records user keys; eg, the REFN tag.

If people put their own info between the @@'s for internal pointers, they
do so at their risk.  I don't know if other programs do as LL does, but I'd
would expect that many do.

Tom Wetmore, ttw@beltway.att.com



          Tuesday, November 8, 1994 7:00:50 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        external links
  To:             GenWeb
I think, the ALIA-lines are a good proposal.
Using this, we could install a first version of a genweb between our Demos.

In my case, I will remove all the files in the genpdb- and genmdb-Directories.
Instead, you find people via:

http://faui80.informatik.uni-erlangen.de/cgi-bin/ll/BASE=/LANG=/F=/N=/?Lookup

( is either germ, engl or dutch) 
If the person has several names, I use a @ to separate them. (Is the _ better?)

If you know the internal index, you may access via:

http:///faui80.informatik.uni-erlangen.de/cgi-bin/ll/BASE=/LANG=/INDEX=I/?LookupInternal

At present, I run the data bases: kaiser, koenige, habsburg, kdg, kdgengl,
kdgdutch.

If we come to a more general adressing scheme, this can be changed quickly.



          Tuesday, November 8, 1994 7:00:58 AM
          GenWeb Item
  From:           baud@research.att.com,Internet
  Subject:        Re:  external links
  To:             GenWeb
> If people put their own info between the @@'s for internal pointers, they
> do so at their risk.  I don't know if other programs do as LL does, but I'd
> would expect that many do.

PAF also manipulates the text used for xref's at it's discretion.
kurt :-)



          Tuesday, November 8, 1994 10:24:35 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        The net is created!
  To:             GenWeb
I made the first 5 connections between erlangen and hull!

If you enter

http://faui80.informatik.uni-erlangen.de/html/ww-person.html

and go via the classification index

to 

K"onige deutscher L"ander

and there to

Index-Koenig-Hannover

you get 5 links to Brian Thompsett in Hull:

Click the person, click again at the personpage: Mehr Information
(more info) -- and you are in Hull in Brians data base!

Everything is generated out of lifelines!!



          Tuesday, November 8, 1994 10:50:03 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        net standardization
  To:             GenWeb
Now, as the possibility is proved, should we start to standardize our procedures.

I got the impression, we all programmed the same stuff.

We should try to develop a list of access procedures = a common genweb interface.

The genweb-urls could be like:

/cgi-bin//

the access info is www-adress-dependent.

What do we need as procedures?

- person access
- name search
- index
- ....



          Wednesday, November 9, 1994 10:03:42 AM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@ucsd.edu,Internet
  Subject:        Building the GenWeb
  To:             GenWeb
While others have been working on the tools to create the GenWeb databases
that will publish genealogical data in the World Wide Web format, I have
done some work at the network level to help establish a common network. We
are taking the first baby steps here and are bound to make some mistakes,
so rather than yelling at the baby, please help steady the wobbling walker.

I have requested assignment of the Internet domain name "genweb.org" to the
as-yet-to-be-formed non-profit organization called The GenWeb Foundation.
The network managers at University of California San Diego have agreed to
provide domain name services for this domain. I propose that each database
in the GenWeb assume a host name in the genweb.org domain. (This will be an
alias; all other names will still be valid.) For example, we could have
"stoyan.genweb.org" and the domain name server will point to the IP address
of Herb Stoyan's server. Further, I propose that one server on the network,
perhaps owned by GenWeb Foundation (assuming we can fund it), be designated
"index.genweb.org" which will keep track of all the other participating
GenWeb servers. Specifically, the index.genweb.org machine will keep track
of the index.html page on each server. This way, there will be no need to
keep a global index of names, but rather an index of each GenWeb server's
index.

Regarding the formation of The GenWeb Foundation, some of us in California
are prepared to file incorporation papers to establish this as a non-profit
organization under California laws that will assist in its becoming an
educational, tax-exempt organization that can accept donations of funds and
equipment. The purposes of The GenWeb Foundation will be as described to
the domain name registration authority:

    "The Genweb Foundation is a non-profit educational organization that
supports genealogy research by defining standards for maintaining
computerized genealogy databases compatible with World Wide Web
communications techniques. The Genweb Foundation conducts educational programs
for researchers in many fields in the use of networked genealogy databases
and coordinates with other network database organizations to maintain data
integrity and compatibility."

In the spirit of the Internet, this will not be an enforcement agency, but
rather a standards-setting  group run on democratic principles. I solicit
your support for this proposal and encourage you to volunteer for a seat on
the board of directors of The GenWeb Foundation. California residence is
not required. (You can even be an illegal alien....)

You may respond either to the list (genweb@ucsd.edu) or to me personally.


Cheers,
Gary



***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************


          Wednesday, November 9, 1994 10:03:42 AM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@ucsd.edu,Internet
  Subject:        Mosaic NetScape is Hot
  To:             GenWeb
I have found that Mosaic Netscape, from the folks who orginally wrote NCSA
Mosaic but are now in a private venture called Mosaic Communications in
Silicon Valley, is a very speedy browser. I recommend that anyone using
NCSA Mosaic try this out. There are versions for Mac, Windows, and
X-Windows. Part of the speed increase is that the browser seems to
establish multiple connections to pick up pictures and other files from an
HTML page.

Cheers,
Gary

***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************


          Wednesday, November 9, 1994 10:57:20 AM
          GenWeb Item
  From:           Joseph D. Barrus,barrus@nosc.mil,Internet
  Subject:        Re: Mosaic NetScape is Hot
  To:             GenWeb
>I have found that Mosaic Netscape, from the folks who orginally wrote NCSA
>Mosaic but are now in a private venture called Mosaic Communications in
>Silicon Valley, is a very speedy browser. I recommend that anyone using
>NCSA Mosaic try this out. There are versions for Mac, Windows, and
>X-Windows. Part of the speed increase is that the browser seems to
>establish multiple connections to pick up pictures and other files from an
>HTML page.
>
>Cheers,
>Gary
>


This is very true, but it also puts a great demand on the network.  Our site
is considering banning its use and some  pages are preventing access from
this client.  A  page on our site has already done this.  Imagine if
everybody used Netscape and were connecting to the same page, especially a
page with great demand.  

Joe
Joseph D. Barrus
NCCOSC RDT&E DIV
barrus@nosc.mil



          Wednesday, November 9, 1994 1:43:32 PM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Re:  Building the GenWeb
  To:             GenWeb
Gary Hoffman writes:
>The network managers at University of California San Diego have agreed to
>provide domain name services for this domain. I propose that each database
>in the GenWeb assume a host name in the genweb.org domain. (This will be an
>alias; all other names will still be valid.) For example, we could have
>"stoyan.genweb.org" and the domain name server will point to the IP address
>of Herb Stoyan's server.

I used "genweb.org" in this way in my earlier example.  Now that we
seem to be approaching implementation, let's be careful about the
details.  We don't know yet how large the list of databases will be,
what the names will look like, or what other services the GenWeb
Foundation may want to catalogue in the DNS.  Therefore I'd like to
reserve a subdomain, say "db.genweb.org", specifically for the list
of Web databases across the Internet.  Thus "stoyan.db.genweb.org"
would be the alias instead.  The "db" subdomain would not be used
for other services, such as "index.genweb.org", so there would be
no risk of name conflicts between databases and other services.

Having a separate subdomain also allows for experiments with alternate
DNS server configurations for the database names without involving
the NIC each time, if we would like to do that in the future.

Second (sorry for getting technical here), you seem to suggest using
DNS "A" records for storing numeric IP addresses of other peoples'
computers (maybe you don't).  I don't think that is a good idea,
since you would then need to update your information also when some
server manager changes the host's IP address (but not its name).  Use
DNS "CNAME" records instead, pointing at the canonical domain name.
(This may be merely a matter of what we mean by "IP address".)

Please keep in mind that all this is still an experiment, and not
something which will be cast in stone and enforced for the forseeable
future.  I have yet other ideas for how to map GEDCOM cross-reference
identifiers to WWW resources.  One is to use the DNS "TXT" record to
store an entire URL rather than just the domain name, and that would
take care of some of the problems mentioned earlier.  However, that
requires more work on the GenWeb servers before it can even be tried.

> Further, I propose that one server on the network,
>perhaps owned by GenWeb Foundation (assuming we can fund it), be designated
>"index.genweb.org" which will keep track of all the other participating
>GenWeb servers. Specifically, the index.genweb.org machine will keep track
>of the index.html page on each server. This way, there will be no need to
>keep a global index of names, but rather an index of each GenWeb server's
>index.

You will also need a clearinghouse for database names providing the
information needed to build the DB.GenWeb.Org domain (maybe this is
included in your "index" service above), unless the network managers
at UCSD have agreed to do it as part of their domain name service
(I believe they have not).  You will probably want to maintain more
information about the databases than is visible in the DNS records.

>    "The Genweb Foundation is a non-profit educational organization that
>supports genealogy research by defining standards for maintaining
>computerized genealogy databases compatible with World Wide Web
>communications techniques. The Genweb Foundation conducts educational programs
>for researchers in many fields in the use of networked genealogy databases
>and coordinates with other network database organizations to maintain data
>integrity and compatibility."

Already the name "GenWeb" seems to imply a connection to *the*
World-Wide Web as we know it, but do we really want to stress
that connection in such a formal statement?  I don't think we
should limit ourselves to any particular standard or technology,
although for the moment the WWW is more than enough to handle.
I'd suggest replacing that reference with a more general one,
such as "distributed computerized genealogy databases" (period).
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE



          Thursday, November 10, 1994 9:07:51 AM
          GenWeb Item
  From:           Brian
Tompsett,B.C.Tompsett@computer-science.hull.ac.uk,Internet
  Subject:        The net is created!
  To:             GenWeb
Hertbert wrote:
> I made the first 5 connections between erlangen and hull!

Just to note that I havn't yet taken my own advice and made my own
record pointers "longlived". They are still byte offsets into the data file
ala Abell-web. I shall be rewriting to remove this feature shortly and
keep people informed when the links might expire into wild pointers.

  Brian


          Thursday, November 10, 1994 9:07:55 AM
          GenWeb Item
  From:           Brian
Tompsett,B.C.Tompsett@computer-science.hull.ac.uk,Internet
  Subject:        Merging Records (was Re: External Links)
  To:             GenWeb
Scott writes:
> [...] The standard goes on to say that when you confirm the suspicion you
> should merge the two records -- possibly good advice for all of
> those Queen Victoria records.

 I don't agree with this in the context of a widely distributed genealogical
database. I think it is better to have a variety of records that point to each
other than a centralisation of data. You avoid problems with single points of
failure, distribute the workload around the network and so on. It also allows
Herbert to maintain a database of Hanovarians in German and me one in English
without having to decide which should be the language or format to standardise
on.

  Brian


          Friday, November 11, 1994 8:06:47 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        Index available
  To:             GenWeb
I have installed an index which looks in all my data bases.

Use the URL

http://faui80.informatik.uni-erlangen.de/cgi-bin/genweb-index/F=/N=

 may be empty or a sequence of name parts separated by @ (don't use
blanks). Multiple name parts make sense for first names only.  A name part
is used for a substring-search. Use * as wildcard character. The exact sequence
of name parts does not matter.

Please report any problems.



          Sunday, November 13, 1994 3:14:44 PM
          GenWeb Item
  From:           Mike Patton€ genealogical
correspondence,MAP=Genealogy@BBN.COM,Internet
  Subject:        Naming and locating info in the GenWeb
  To:             GenWeb
Since Genealogy is my hobby, not a profession, it sometimes takes a
back seat to other things...  I'm replying to several messages from
the last two weeks...they were interrelated, so I'm including them in
one reply.  The total message is pretty long...  Sections quoted from
each of the original messages start with identifying headers.  Also,
there are some more general comments on this thread at the end...

   Date: Wed, 2 Nov 1994 17:43:19 -0700
   From: Bill Harten 

   Please let me introduce myself.

OK, I'll take this opportunity to introduce myself as well, as I think
it's relevant...  I am a Network Engineer/Analyst at BBN, doing
operational engineering for a world-wide IP network (the DSI) which
includes setup and supervision of DNS and HTTP servers (DNS for
DSI.NET domain [and I previously ran the LCS.MIT.EDU domain, cited in
many examples in the DNS standard] and web server http://www.dsi.net).
I am a member of the IETF (the Internet Engineering Task Force that
sets Internet standards), and have served as the co-chair of the DNS
working group.  Computers and networking are also a hobby (I love it
when they pay me to do stuff I'd pick as a hobby :-) and I have seven
or eight computers at home (excluding the ones in my microwave,
toaster, clock, etc. the exact number depends on how you count the
collected "parts" some of which are sufficient for more machines I
just haven't gotten to assembling, yet :-).  My house will soon have a
full time 2Mbps connection to the Internet (assuming it all works
out), and I will be putting my data on the Web from there.

I've been gathering my personal genealogy info for about 5 years now.
Both of my parents come from large extended families, both of which
hold regular reunions (Pattons every two years in NE KY, about 30-40
people [four generations down]; Goffs every year in SE MA, close to
1000 people [eight to nine generations down] :-).  The "Goff Gathering
Association" already had an official genealogist, but the Pattons only
had word-of-mouth and reminiscence.  About 5 years ago, during a
reminiscence session at the reunion, one of the older relatives (a
second cousin once removed) sketched a tree of a part of the family I
had not known about.  This inspired me to start capturing some of the
info.  It was all "downhill" from there...

   I have been watching GenWeb from the beginning with intense interest, and 
   believe the basic concept to be crucial to the future of genealogy.

Yes.  When I saw the GenWeb announcement, I thought it had great
potential.  Since I have some professional credentials in the area, I
decided to join the effort, even though I don't have an awful lot of
spare time right now (and thus the two week delay in responses :-).

   ... from Herbert Stoyan:

   >Additionally, gedcom lacks the following things:
   >...
   >external-db-links (for people which are mentioned in one db but the full 
   >info>is in another db)

   External references are critical.  There has to be some way to refer to a 
   record in another database by means of a _permanent_ record identifier, 
   combined with a permanent resource identifier, which could be thought of as 
   a kind of ISBN (don't know the acronym) book identifier.  The 
   resources/records thus identified have to be able to move around as systems 
   come and go, without invalidating pointers in other databases/pages.

What you describe is exactly what the URN group in the IETF is
designing.  Persistant "names" for objects that can move around
arbitrarily.  Like URLs, they start with a "context", one of those is
"isbn:", another could easily be "genweb:" ... we'd just have to
decide on the format...

   From my understanding of the web's URL format specifications, I now see that 
   the pound sign(#) could be used instead of the colon--URL's had not been 
   defined at the time this concept became part of our architecture.

As someone else pointed out, "#" is not the right thing, but "?" in a
URL can do what you want.  Still URNs are much better since URLs are
NOT persistant.  It may not be necessary to change the GEDCOM spec to
fit this.  If we think that it's general enough (see below), we can
use it as is (i.e. if a GEDCOM external reference were "maps-db:GGF",
we could say that maps to the URN "genweb:maps-db:GGF").  I think the
question of what's a sufficiently good reference WITHIN the GenWeb is
interesting, the details of how that fits into the non-GenWeb system
is not really important.  I suggest that we just trust that it will...

   Date: Thu, 3 Nov 1994 07:27:44 -0500
   From: Gene Stark 

   [I] think that a model similar to [the DNS] might be the most
   reasonable.  ... Anders Andersson ... suggested the use of the
   DNS service itself as an interim solution until a better one is
   invented, but I doubt if the current DNS software can be "bent" far
   enough to be useful in GenWeb, as it was designed with a somewhat
   different application in mind.

Actually, the DNS can easily be bent to provide a really good interim
solution.  The design of URNs is supposed to address the long-term
problem, both more generally and more directly.  I would suggest not
distracting the GenWeb project's efforts away from genealogy, let the
IETF working groups on information resources handle this for us (if
you care about the details, you can easily join the relevant groups).
Building and deploying yet another system for this aspect of the
problem is a lot of work, and I don't see any reason why we should
duplicate this effort.

   [describes detailed scheme]

   An important aspect of this scheme is the separation of the name
   service from the actual Web data.  The actual authoritative
   information about the URL corresponding to a particular resource
   would be known only to the individual servers near where the data
   is stored.  ...  Web documents would still contain URL's but but
   failure to contact a URL would trigger a query to a server to
   determine if the data has moved.

What you describe is EXACTLY what URNs are supposed to do, they are
designed as a general naming service and can easily be fit into
genealogical data as you describe.

   ... One thing I forsee here is that there is likely to be only a
   few "archives" of data on any given individual, but many more
   "mirrors", which are people who have copied this data into their
   own local database.  There should be a way to distinguish between
   "archives" and "mirrors", and have a query propagate only toward
   the archive servers for that individual.  Also, maybe there should
   be a way that access to a mirror automatically triggers a check to
   see if the archive copy has been updated.

URNs can and will do all of these things as well...

   From: Gary Hoffman 
   Date: Wed, 09 Nov 1994 09:54:16 PDT

   I have requested assignment of the Internet domain name "genweb.org" ...

You beat me to it :-).  I was going to suggest doing this in response
to the discussion quoted above (and the rest of the thread).  I would
be glad to offer my DNS expertise in the set up of this and am willing
to run the primary or a secondary out of my house when the 2Mbps link
gets in, giving us bi-coastal US coverage.  A secondary in a
reasonably well connected spot in Europe would probably be good, too.

   I propose that each database in the GenWeb assume a host name in
   the genweb.org domain. (This will be an alias; all other names will
   still be valid.) For example, we could have "stoyan.genweb.org" and
   the domain name server will point to the IP address of Herb
   Stoyan's server.

Actually, CNAMEs (real aliases) would be better.  I.e. when you looked
up "stoyan.genweb.org", what you get back is the answer that it's an
alias for faui80.informatik.uni-erlangen.de which your system would
then look up to find the address.  Thus, if that host moves around in
Germany, the server in Sand Diego doesn't care.

   Further, I propose that one server on the network,
   perhaps owned by GenWeb Foundation (assuming we can fund it), be
   designated "index.genweb.org" which will keep track of all the
   other participating GenWeb servers.

Actually, I'd like to discuss the structure of the genweb.org domain a
bit more, I don't think your examples will give the flexibility
needed.  I agree with another writer who said perhaps the right thing
is a second level.  If the GenWeb Foundation is actually going to own
machines, they'll want to have a piece of the namespace reserved for
that, then we may want to have other parts reserved for other uses.
The setup that seems to have developed as the most natural for most
people is naming machines actually owned by the organization directly
at the top, and adding subdomains for "service groupings" (at MIT-LCS
we had this kind of setup, e.g. .LCS.MIT.EDU for a host and
.LPD-SPOOLER.LCS.MIT.EDU for a printer was a CNAME to the
host that spooled it).  For GenWeb, the groupings might eventually
include actual databases, indexes in various forms and functionality,
and other useful references.

   Specifically, the index.genweb.org machine

I've found it extremely helpful to never give a machine a primary name
which is the service it runs.  This causes innumerable confusions
later when things are changed.  I prefer naming pieces of hardware out
of some arbitrary name space (my machines at home use the names of
starships that the original Star Trek had designated, but never used
on the show; at LCS I used planets from the Foundation books :-).
Then make aliases for use in advertising services.  This makes it
easier to move services around and to upgrade a service machine, but
keep the older machine as backup and providing some lesser service,
without unduly confusing people.  See RFC1178 for more on this topic.

       "The Genweb Foundation is a non-profit educational organization
   that supports genealogy research by defining standards for
   maintaining computerized genealogy databases compatible with World
   Wide Web communications techniques. The Genweb Foundation conducts
   educational programs for researchers in many fields in the use of
   networked genealogy databases and coordinates with other network
   database organizations to maintain data integrity and
   compatibility."

I have a few minor problems with that statement, most notably about
limiting it to a specific technology (WWW) of distribution, but in
general it looks like the right idea.  I'm not, however, convinced at
this point that a legal non-profit is really needed, rather than just
the loose coalition of the members of this list...

   ... encourage you to volunteer for a seat on the board of directors
   of The GenWeb Foundation.

If the Foundation really does form, I'd be glad to offer my services
in whatever way will help best...


----------------------------------------------------------------

General comments...

I think that our best use of energy is to concentrate on the
genealogical aspects of what a globally interconnected genealogy web
would (or should) look like.  I think many of the naming and locating
problems people are discussing are already being adequately addressed
by existing groups and that they will likely have a solution faster
than we could design our own, and it'll probably be better (having
been hashed out for several years) and have more widespread
availability.  The only aspect of naming that I think deserves
attention is how to name individual databases WITHIN the GenWeb, and
not to worry at this time which higher level naming structure it comes
under.  Designing a scheme compatible with both DNS and URNs, and
likely any future ones as well, is the trivial part.  Making it scale
internally may well be the difficult part...  Building prototypes
within DNS is probably a good start on learning what needs to be
addressed.

I think the more important work lies in two areas.  One is designing
reasonably consistent presentation (so as I wander the Web from DB to
DB, I don't have to keep learning new places to look for the links I'm
interested in).  The other is designing an architecture for indexing
and lookup (the work on centroids in the IETF info resources groups is
probably a good place to start here) that can deal with the potential
information explosion.  Think about having all the information
collected by anyone on line and then ask yourself about structures
that would support the kind of questions you are working on in your
own research (for me, the current question of interest is "My
Great-Granfather is reported in several sources as having been born in
Breathitt County, Ky on March 5, 1857, but a courthouse fire destroyed
the records from that time.  I want to find any data on Pattons in
that area [Ky counties were kind of fluid back then, Breathitt didn't
exist in the 1850 census (I think) and parts of what were Breathitt in
1857 had formed other counties by the 1860 census] during that
timeframe.")

[work signature included for information only]

            __
  /|  /|  /|  \     Michael A. Patton, Network Analyst
 / | / | /_|__/     Defense Simulation Internet
/  |/  |/  |atton   Bolt Beranek and Newman

Disclaimer: The opinions expressed above are a figment of the phosphor
on your screen and do not represent the views of BBN, DSI, or MAP. :-)



          Monday, November 14, 1994 10:27:38 PM
          GenWeb Item
  From:           Chris Garrigues,cwg@DeepEddy.Com,Internet
  Subject:        Re: Naming and locating info in the GenWeb
  To:             GenWeb
At  6:07 PM 11/13/94 -0500, Mike Patton wrote:
>Since Genealogy is my hobby, not a profession, it sometimes takes a
>back seat to other things...  I'm replying to several messages from
>the last two weeks...they were interrelated, so I'm including them in
>one reply.  The total message is pretty long...  Sections quoted from
>each of the original messages start with identifying headers.  Also,
>there are some more general comments on this thread at the end...

I've also fallen behind on genweb reading, so I thought yor message was as
good to follow up to as any.

>   Date: Wed, 2 Nov 1994 17:43:19 -0700
>   From: Bill Harten 
>
>   Please let me introduce myself.
>
>OK, I'll take this opportunity to introduce myself as well, as I think
>it's relevant...  I am a Network Engineer/Analyst at BBN, doing

Okay, if we're doing intrductions, I'm a Network/Systems programmer at MCC
taking care of the research environment.  I'm also doing some consulting on
the side for a local ISDN provider.  DeepEddy.Com resides in my spare
bedroom, and I've been neglecting genweb lately because I've been trying to
build a stable environment here.  (What?  I have to do my OWN backups?!?!?)
About two weeks ago, I got a web server running, last week I got automated
backups finally up, and tonight I finally got Lifelines up and running.

>What you describe is exactly what the URN group in the IETF is
>designing.  Persistant "names" for objects that can move around
>arbitrarily.  Like URLs, they start with a "context", one of those is
>"isbn:", another could easily be "genweb:" ... we'd just have to
>decide on the format...

Are you tracking what they're doing?  Our hands are somewhat tied until
they finish.  If you are tracking them, do you know how far they are from
having something useful?

>   From: Gary Hoffman 
>   Date: Wed, 09 Nov 1994 09:54:16 PDT
>
>   I have requested assignment of the Internet domain name "genweb.org" ...
>
>You beat me to it :-).  I was going to suggest doing this in response
>to the discussion quoted above (and the rest of the thread).  I would
>be glad to offer my DNS expertise in the set up of this and am willing
>to run the primary or a secondary out of my house when the 2Mbps link
>gets in, giving us bi-coastal US coverage.  A secondary in a
>reasonably well connected spot in Europe would probably be good, too.

Since my site has stabilized (with a 128KB dial-on-demand) link, I'm
willing to be a secondary (or primary) as well.  I also have DNS expertise.
When you requested genweb.org, did you give them two servers?  Should I
(or Mike or someone else) create a minimal domain with a couple of cnames
for the first 2 or 3 DNS servers and to those hosts which have prototypes
running on them today.

Mike, what kind of home link do you have right now?  If you're going to
have a 2Mbps link, then you really should be the primary, but what have we
got in the short term?

>Actually, CNAMEs (real aliases) would be better.  I.e. when you looked
>up "stoyan.genweb.org", what you get back is the answer that it's an
>alias for faui80.informatik.uni-erlangen.de which your system would
>then look up to find the address.  Thus, if that host moves around in
>Germany, the server in Sand Diego doesn't care.

Agreed, 100%.

>   Further, I propose that one server on the network,
>   perhaps owned by GenWeb Foundation (assuming we can fund it), be
>   designated "index.genweb.org" which will keep track of all the
>   other participating GenWeb servers.
>
>Actually, I'd like to discuss the structure of the genweb.org domain a
>bit more, I don't think your examples will give the flexibility
>needed.  I agree with another writer who said perhaps the right thing
>is a second level.  If the GenWeb Foundation is actually going to own
>machines, they'll want to have a piece of the namespace reserved for
>that, then we may want to have other parts reserved for other uses.
>The setup that seems to have developed as the most natural for most
>people is naming machines actually owned by the organization directly
>at the top, and adding subdomains for "service groupings" (at MIT-LCS
>we had this kind of setup, e.g. .LCS.MIT.EDU for a host and
>.LPD-SPOOLER.LCS.MIT.EDU for a printer was a CNAME to the
>host that spooled it).  For GenWeb, the groupings might eventually
>include actual databases, indexes in various forms and functionality,
>and other useful references.

A lot of this depends on what the GenWeb Foundation turns out to be, but
you will never go wrong by designing so that you can scale.  That's why my
Sun has several cnames pointing to it even though it's only one machine.

>   Specifically, the index.genweb.org machine
>
>I've found it extremely helpful to never give a machine a primary name
>which is the service it runs.  This causes innumerable confusions
>later when things are changed.  I prefer naming pieces of hardware out
>of some arbitrary name space (my machines at home use the names of
>starships that the original Star Trek had designated, but never used
>on the show; at LCS I used planets from the Foundation books :-).
>Then make aliases for use in advertising services.  This makes it
>easier to move services around and to upgrade a service machine, but
>keep the older machine as backup and providing some lesser service,
>without unduly confusing people.  See RFC1178 for more on this topic.

Agreed again.

>       "The Genweb Foundation is a non-profit educational organization
>   that supports genealogy research by defining standards for
>   maintaining computerized genealogy databases compatible with World
>   Wide Web communications techniques. The Genweb Foundation conducts
>   educational programs for researchers in many fields in the use of
>   networked genealogy databases and coordinates with other network
>   database organizations to maintain data integrity and
>   compatibility."
>
>I have a few minor problems with that statement, most notably about
>limiting it to a specific technology (WWW) of distribution, but in
>general it looks like the right idea.  I'm not, however, convinced at
>this point that a legal non-profit is really needed, rather than just
>the loose coalition of the members of this list...

wouldn't that be "genweb.disorg"?  ;-)

Seriously, I'd like to hear arguments for why the GenWeb Foundation should
be a real legal entity.  Would the benefits ouweigh the costs, or would it
be more effective to run it as an adhocracy, borrowing time and resources
from the various people who are willing to provide it (which is what it is
right now).

>   ... encourage you to volunteer for a seat on the board of directors
>   of The GenWeb Foundation.
>
>If the Foundation really does form, I'd be glad to offer my services
>in whatever way will help best...

If we see that a real Foundation needs to exist, I'm willing to be
involved.  If it continues to run under the radar, I'm also willing to be
involved.

>I think the more important work lies in two areas.  One is designing
>reasonably consistent presentation (so as I wander the Web from DB to
>DB, I don't have to keep learning new places to look for the links I'm
>interested in).  The other is designing an architecture for indexing
>and lookup (the work on centroids in the IETF info resources groups is
>probably a good place to start here) that can deal with the potential
>information explosion.  Think about having all the information
>collected by anyone on line and then ask yourself about structures
>that would support the kind of questions you are working on in your
>own research (for me, the current question of interest is "My
>Great-Granfather is reported in several sources as having been born in
>Breathitt County, Ky on March 5, 1857, but a courthouse fire destroyed
>the records from that time.  I want to find any data on Pattons in
>that area [Ky counties were kind of fluid back then, Breathitt didn't
>exist in the 1850 census (I think) and parts of what were Breathitt in
>1857 had formed other counties by the 1860 census] during that
>timeframe.")

I'm really interested in the indexing issues.  The presentation issue is
also important, but since it interests me less, I'll probably pretty much
follow whatever scheme others want to standardize on.  Those who are
interested in presentation issues should probably start by deciding just
how much look&feel *should* be specified.

Where can I find the centroid references?  I'd like to read it in my
copious spare time.

Chris


--
Chris Garrigues
cwg@DeepEddy.Com
My homepage



          Tuesday, November 15, 1994 9:23:43 AM
          GenWeb Item
  From:           Chris Garrigues,cwg@DeepEddy.Com,Internet
  Subject:        Re: Naming and locating info in the GenWeb
  To:             GenWeb
At  9:35 AM 11/15/94 +0100, Anders Andersson wrote:
>Chris Garrigues writes:
>>When you requested genweb.org, did you give them two servers?  Should I
>>(or Mike or someone else) create a minimal domain with a couple of cnames
>>for the first 2 or 3 DNS servers and to those hosts which have prototypes
>>running on them today.
>
>You can have CNAME records pointing at the GenWeb database servers,
>but the NS records may not refer to alias names, only to real host
>names, if I have understood the DNS RFC's and the discussions on
>comp.protocols.tcp-ip.domains correctly.  Therefore, we need to pick
>the official name servers with some care before filing them with the
>NIC (of course they can be changed later by a new request to the NIC,
>but that's not quite as simple as changing a CNAME record in our own
>domain).

You're right, you do want to pick NS servers which are unlikely to change.
I also think that as distributed as we are, we should have at least 3
servers rather than the required two registered.

>If someone sets up a minimal GENWEB.ORG zone as SOA first and informs
>the volunteer NS maintainers (of which I represent one) about its
>address, we can check that it all works fine before picking the
>official NS hosts and telling the NIC about them.  Mike?

Okay, there's a minimal zone on ns1.deepeddy.com (192.195.85.8) for
genweb.org; it contains an SOA, a single NS, and two CNAMES.

I should point out that since this is a dial-on-demand ISDN line, the first
time you ping it, it'll take about half a second to bring up the line, then
the connection is plenty fast.  The line drops after 30 seconds.  (I do't
actually have to pay for my connection because of my "connections", but I
try to be a good citizen, and the half second rampup time is generally
sufficient.)

Chris


--
Chris Garrigues
cwg@DeepEddy.Com
My homepage



          Wednesday, November 16, 1994 12:06:30 AM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@ucsd.edu,Internet
  Subject:        Why The GenWeb Foundation?
  To:             GenWeb
I am very pleased that we have attracted some "heavy hitters" to this
mailing list who are part of the very foundation of the Internet. I
appreciate all the comments about setting up the domains needed to run
GenWeb and will gladly defer to others' expertise in this area.

Some have asked whether we need a formal organization. That is a good
question. My own vision of GenWeb is that it will be a loose affiliation of
servers around the world run by volunteers. My fear is that without a
standards-setting group, we won't achieve the uniformity that allows
seamless server-to-server links as users climb their pedigrees. Besides
standards-setting (the term sends a chill up the spine of any Internet
frontiersman) is the concern for long-term data integrity.

Any volunteer can just as easily unvolunteer, leaving a gap in the radar
coverage (to borrow a metaphor). This would especially impact any host
acting in a key position, such as an index server. I believe we can count
on the longer-gevity of an organization vis-a-vis an individual. For
example, can we rely on a major GenWeb server that is actually running
"under the radar" of its actual owner, some university or corporation? If
the employer has put large amounts of GenWeb data in his WWW box and then
is laid off, where is the integrity of our network? Should we establish
redundant "under the radar" servers to protect against this eventuality? I
am a good example of this as I anticipate leaving my position here (and my
high-speed Internet connection) after I graduate from law school. I'm still
trying to find a home for my GenWeb server on campus so that it won't get
stranded. But, in a sense, we are all "between job searches."

Bill Minnick had suggested to me that we might form a non-profit org that
could assess "dues" from large family organizations and thus create some
capital to be used for network coordination functions. I was at first
hesitant in even moving to formalize our loose-knit convocation, but I see
some virtue in the suggestion. In fact, I even used the name when I
registered GWNWEB.ORG with the InterNic. (BTW, I do not have confirmation
back from domreg yet.) It did seem to give my request more weight, but I
also understand that they will register anything that is not already
registered and is not obscene.

I have further reasons for recommending a formal organization. As a law
student, I have become sensitive to issues of privacy, copyright, and
defamation, all of which a GenWeb server can run afoul of thanks to data
being uploaded from well- (or ill-)meaning contributors. In the process,
the owner of the host machine (i.e. your employer) may find a lawsuit on
its hands thanks to the data in your machine which is being published to
the world. As an exercise in my Computer Law class, I have drafted a
Contributor-Sysop and a User-Sysop agreement which I will post to the list
very soon. These agreements try to deflect liability from the GenWeb Sysop.
I believe they represent something that, in a final form, should be adopted
by a standards-setting group that has general support among us all.
Otherwise, we might all be found collectively liable for tolerating
maverick operations.

I hope you don't think that the fun is over when the lawyers arrive. (I am
not a lawyer, I just play one on the Internet.) The idea is that a good
contract or agreement is like carrying an umbrella: just hope it doesn't
rain, but you are prepared if it does.

Please speak up on these issues.

Cheers,
Gary


***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************



          Wednesday, November 16, 1994 1:03:28 PM
          GenWeb Item
  From:           Everton,jayhall@xmission.com,Internet
  Subject:        Update to Web page: Genealogy on the World-Wide Web
  To:             GenWeb
 
Everton Publisher's WWW page has added a new area, "Personal Views of 
Genealogy on the Internet", featuring links to personal Web pages with 
genealogical themes. This area is available via Everton's front page:
 
                     http://www.xmission.com/~jayhall/
 
Everton's Web pages contain links to over 200 files, Web pages, gopher 
menus, and databases. Current areas include: Getting Started With Your 
Genealogical Research, Genealogical Resources in the United States, 
Non-U.S. Genealogical Resources, Special Genealogical Resources, Using the 
Internet, Personal Views of Genealogy on the Internet, and Everton 
Publishers, with other areas currently under construction.
 
--------------------------------------------------------------------------
Jay Hall, Everton Publishers     | WWW = http://www.xmission.com/~jayhall/
165 South West Temple, Suite 200 | E-Mail = jayhall@xmission.com
Salt Lake City, UT 84101         |



          Sunday, November 20, 1994 11:04:43 PM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@ucsd.edu,Internet
  Subject:        Registration of GENWEB.ORG
  To:             GenWeb
I have finally received confirmation from the Domain Name authority that my
request to register GENWEB.ORG has been approved. (The confirming message
is appended below.) This will put us on the map, but more importantly, it
gives us a unified naming scheme for the GenWeb network.
Is anyone ready to have their GenWeb database listed with a "cname" (i.e.
alias) in the GENWEB.ORG domain? If so, please post a notice to this
mailing list. Here is the information we need:

Requested machine name (alias):
IP number of this machine:
Machine type and Operating System:
Name of primary operator:
e-Mail address of primary operator:
Telephone number of primary operator:
Path to GenWeb index page on your system:

Here is the text of the confirmation message, dated 11/18/94:

 Registration for the domain GENWEB.ORG has been completed.
 The InterNIC database has been updated; see the record below.

 Please let us know if any changes need to be made.

 The new information will not be visible via WHOIS until the next
 business day.

 Before this information is released to the Internet (normally
 accomplished via root server file updates on Mon, Wed, and Fri
 afternoons about 17:00 Eastern time) we will check to see that the 
 primary and secondary name servers are reachable and contain the 
 appropriate SOA and NS resource records.

 If your name servers are not properly configured, we will place your
 registration on hold until we hear from you.  When you are ready to 
 have the domain released to the network, contact us at any of
 the addresses below.

 Registering a domain name does not confer any legal rights to that
 name and any disputes between parties over the rights to use a
 particular name are to be settled between the contending parties
 using normal legal methods (See RFC 1591).

   InterNIC Domain Registration

   email  domreg@internic.net       telnet, ftp, whois, gopher, wais
   voice  (703) 742-4777               rs.internic.net or 198.41.0.5
   fax    (703) 742-4811            nslookup, dig
                                       ns.internic.net or 198.41.0.4

 ====================================================================

The Genweb Foundation. (GENWEB-DOM)
   3950 La Jolla Village Dr.
   La Jolla, CA 92037

   Domain Name: GENWEB.ORG

   Administrative Contact, Technical Contact, Zone Contact:
      Hoffman, Gary   (GH126)  ghoffman@UCSD.EDU
      (619) 534-7733

   Record last updated on 18-Nov-94

   Domain servers in listed order:

   UCSD.EDU                     128.54.16.1 132.239.1.1 132.239.254.201
   NETWORK.UCSD.EDU             128.54.16.3, 132.239.254.203




***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************



          Tuesday, November 22, 1994 7:08:38 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        genweb.org
  To:             GenWeb
alias of the machine: www8.informatik.uni-erlangen.de
IP number: 131.188.38.1
machine type and os: SUN, Solaris
primary operator: Herbert Stoyan
e-mail: hstoyan@informatik.uni-erlangen.de
phone: 0049-9131-699-130
path to genweb index: entry-pages are: html/chl-enter, html/ww-person.html,
html/parcham.html

to get an index, use url:
http:www8.informatik.uni-erlangen.de/cgi-bin/genweb-index/F=/N=



          Wednesday, November 23, 1994 4:08:31 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        genweb-index
  To:             GenWeb
----------
X-Sun-Data-Type: text
X-Sun-Data-Description: text
X-Sun-Data-Name: text
X-Sun-Content-Lines: 22

I get the impression, that due to our progress, the discussion ha ended.

Is really everything clear now?

I believe, the result of the discussion was, every host should decide on its
own, ho to link to other hosts.
I use now the ALIAS-proposal to realize links to other data bases. 

What we should standardize, is the index access. I see no reason to create
`index pages'. Instead, we should provide index procedures. Even people who
don't believe that names are a good pointer scheme should accept that they
are the basis of a query. Or not?

My proposal is a cgi-bin procedure which delivers an index according to a
question: genweb-index gets parameters F=firstnames N=surnames and delivers a set
of pointers to the person records at the current site.

As you may know, I don't use ged2html. Instead, I use lifeline reports. Here I
add: The genweb-index shell script and the namesearch report.

Namesearch understand a wildcard * and understands sequences of names as being 
to processed via all permutations. Spaces between the name parts are to be
filled with @, which may be not the best proposal.

----------
X-Sun-Data-Type: default-app
X-Sun-Data-Description: default
X-Sun-Data-Name: namesearch
X-Sun-Content-Lines: 146

global(r)
proc main ()
{
        getstr(d, "What is the database?")
        set(db,save(d))
        getstr(l, "What is the firstname?")
        set(i,save(l))
        getstr(k, "What is the surname?")
        set(j,save(k))
 " "
        indiset(people)
  set(starf,index(i,"*",1))
  set(blank,index(i," ",1))
  set(stars,index(j,"*",1))
  set(c1,strlen(i))
  set(c2,strcmp(i,"?"))
  set(c1,strlen(i))
  set(c2,strcmp(i,"?"))
  if(or(eq(0,c1),eq(0,c2))){
        forindi(p,x){set(n,"")
                     set(v,"")
                     if(n,surname(p)){ set(n,save(surname(p)))
                                       set(c,index(n,j,1)) 
                                       if(ne(c,0)){addtoset(people,p,y)}
                                       else{if(ne(stars,0)){
                                                         call starscan(n,j)
                                                         if(eq(r,1)){
                                                     addtoset(people,p,y)}}}}}}
  else{set(c1,strlen(j))
       set(c2,strcmp(j,"?"))
       if(or(eq(0,c1),eq(0,c2))){
        forindi(p,x){set(n,"")
                     set(v,"")
                     if(n,givens(p)){  set(n,save(givens(p)))
                                       set(c,index(n,i,1))
                                       if(ne(c,0)){addtoset(people,p,y)}
                                       else{if(ne(starf,0)){
                                                         call starscan(n,i)
                                                         if(eq(r,1)){
                                                     addtoset(people,p,y)}}}}}}
       else{forindi(p,x){set(n,"")
                         set(v,"")
                         if(n,givens(p)){set(n,save(givens(p)))
                                         set(c,index(n,i,1))
                                         if(ne(c,0)){if(s,surname(p)){
                                                        set(s,save(surname(p)))
                                                        set(c,index(s,j,1))
                                                        if(ne(c,0)){
                                                          addtoset(people,p,y)}
                                                        else{if(ne(stars,0)){
                                                         call starscan(s,j)
                                                         if(eq(r,1)){
                                                     addtoset(people,p,y)}}}}}
                                         else{if(ne(starf,0)){
                                                         call starscan(n,i)
                                                         if(eq(r,1)){
                                                           if(s,surname(p)){
                                                            set(c,index(s,j,1))
                                                            if(ne(c,0)){
                                                          addtoset(people,p,y)}
                                                        else{if(ne(stars,0)){
                                                         call starscan(s,j)
                                                         if(eq(r,1)){
                                                     addtoset(people,p,y)}}}}}
 }}}}}}
       set(no,lengthset(people))
       if(ne(no,0)){
	"\n"
        "\n"
	forindiset(people,s,y,z) {call pperson(s,db)}
        "\n\n"}	
}

proc pperson (i,db)
{
	"
" "" if (t,title(i)) {t " "}
fullname(i,0,1,300) "
\n"
	if (e, birth(i)) {
		"*: " long(e) "
\n"
	}
	if (e, death(i)) {
		"+: " long(e) "\n"
	}
	"
\n"

}

proc starscan(s,m)
{ set(j,1)
  set(r,0)
  set(match,0)
  list(l)
  list(k)
  list(n)
  while(i,index(m,"*",j)){enqueue(l,i)
                          enqueue(k,1)
                          enqueue(n,1)
                          set(j,add(j,1))}
  set(i1,1)
  set(c,1)
  if(ne(getel(l,1),1)){if(ge(getel(l,1),strlen(s))){set(c,0)}else{
    set(s1,save(substring(s,1,sub(getel(l,1),1))))
    set(c,index(s1,substring(m,1,sub(getel(l,1),1)),1))
    if(eq(c,1)){ 
      set(i1,getel(l,1))}}}
  set(match,1)
  if(ne(c,1)){set(match,0)}
  else{
  set(i,1)
  while(le(i,length(l))){
  
if(eq(i,length(l))){set(mlen,strlen(m))}else{set(mlen,sub(getel(l,add(i,1)),1))}
   if(ne(getel(l,i),strlen(m))){
   set(s1,save(substring(s,i1,strlen(s))))
   set(c,index(s1,substring(m,add(getel(l,i),1),mlen),getel(k,i)))
   if(eq(c,0)){decr(i)
               if(eq(i,0)){set(match,0)
                           set(i,add(length(l),1))}
               else{set(i,getel(n,i))
                    setel(k,i,add(1,getel(k,i)))}}
   else{incr(i)
        if(le(i,length(l))){set(i1,add(i1,sub(mlen,getel(l,i))))}}}
   else{incr(i)}}
  set(r,match)}
}

proc blankscan(s,m)
{ set(j,1)
  set(l,1)
  set(c,1)
  while(ne(c,0)){
    set(i,index(m," ",j))
    if(ne(i,0)){
      set(p,save(substring(m,l,sub(i,1))))
      set(c,index(s,p,1))
      if(ne(c,0)){
        set(l,add(i,1))
        incr(j)}
      else{set(d,index(p,"*",1))
           if(ne(d,0)){set(r,0)
                       call starscan(s,p)
                       set(c,r)}}}
    else{set(c,0)}}
  if(eq(i,0)){set(r,1)}else{set(r,0)}
}

----------
X-Sun-Data-Type: shell-script
X-Sun-Data-Description: shell-script
X-Sun-Data-Name: genweb-index
X-Sun-Content-Lines: 24

#!/bin/sh

PATH_INFO=`echo $PATH_INFO|sed -e 'y/@/ /'`
name=`echo $PATH_INFO|sed -e 's/^.*N=\([^\/]*\).*$/\1/'`
fname=`echo $PATH_INFO|sed -e 's/^.*F=\([^\/]*\).*$/\1/'`

        echo "Content-type: text/html"
	echo

# here similar parts follow for all data bases at the host. The actual data 
# base here is "kdg". The directories will differ on other hosts

	mkdir /tmp/ged.$$
	LLDATABASES=/user8c/hstoyan/bocal/lissp/kdg
	LLPROGRAMS=/user8c/hstoyan/bocal
	LLREPORTS=/tmp/ged.$$
	TERM=vt100
	export LLDATABASES LLPROGRAMS LLREPORTS TERM
        (echo r/user8c/hstoyan/bocal/namesearch ; echo kdg ; echo $fname ; echo
$name ; echo /tmp/ged.$$/result.html ; echo q) | /user8c/hstoyan/bocal/lines
/user8c/hstoyan/bocal/lissp/kdg > /tmp/ged.$$/logfile 2>&1 
        cat /tmp/ged.$$/result.html
        cd /
        rm -rf /tmp/ged.$$

	;;


          Wednesday, November 23, 1994 5:55:18 PM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@UCSD.EDU,Internet
  Subject:        Indexing GenWeb Servers
  To:             GenWeb
A user seeking to find a genealogical record in GenWeb space need only
submit a single search request. I have proposed that such a request be
issued to a GenWeb index server (possible on index.genweb.org) that keeps
track of all the known/registered/acknowledged GenWeb servers. I invite
response on what should happen next. For example, should the index server
in turn relay the search request to the various servers who would generate
a response back to the requestor or should the index server merely respond
to the requestor with the list of available servers which the requestor
could then manually or automatically access directly? The actual search
mechanism on each server is irrelevant to this discussion, as long as we
can standardize on a search request protocol.

I propose a search protocol somewhat along the lines that Herb Stoyan is
already implementing. That is, on each server, a search is launched by
sending a www request to:
   http://...path to genweb search
engine.../[familyname],[givenname],[birthyear]
Each argument should be optional, but delimited with commas (assuming
commas are permitted in a URL).
Also some leeway should be given to alternate spellings for names and year
ranges for birthyear.
The search engine should return an HTML document to the user listing all
possible matches to the request. If this cannot be returned before the
user's browser quits waiting, then we need to cache the result somewhere
where it can be retrieved later by the user.

Is this concept achieveable given the state of the art? Other questions:
Should the user get a single report back from all servers or a separate
report from each known server?
How else does this concept lend itself to automation?
Further, does anyone suppose it will be possible for a function on a server
to survey the records on that server to identify missing links (usually
parental) and then to initiate a search on all other servers to try to
establish these links?

Something to think about over Thanksgiving dinner.

Cheers,
Gary

 
***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************


          Wednesday, November 23, 1994 10:09:19 PM
          GenWeb Item
  From:           Mike Patton€ genealogical mail,MAP=Genealogy@BBN.COM,Internet
  Subject:        Re: Indexing GenWeb Servers
  To:             GenWeb
   From: Gary Hoffman 
   Date: Wed, 23 Nov 1994 13:20:01 PDT

   A user seeking to find a genealogical record in GenWeb space need only
   submit a single search request.

This is a very laudable goal, but you need to be really careful about
scaling issues.  While we're only a small group this may work fine,
but if we expect the GenWeb to grow until there are lots of users
(which is neeeded if it's to become truly useful), scaling properties
could cause us to die of success, I've seen it happen to other
projects.  If the GenWeb gets large enough to really pay off, it won't
be possible to have a single index of everything.  Some of the
research in the IETF IIIR and related groups could help here.  They
have several documents in preparation which can be visited from the WG
page http://www.ietf.cnri.reston.va.us/html.charters/iiir-charter.html).

   I have proposed that such a request be issued to a GenWeb index
   server (possible on index.genweb.org) that keeps track ...

This is one of the scaling points that's hard.  If there's only one
server, it could end up fielding more requests than it can handle,
especially if it has to do much work.  A better architecture would be
to have this as a general service that any (or at least many) of the
GenWeb servers can perform.the first-level work.  That way, no one
server gets overloaded.  This is a problem that almost killed off
Archie several times.  I'm going to see if I can get some time with
some of the Bunyip folks week after next at IETF to discuss this with
them.

   ... what should happen next. For example, should the index server
   in turn relay ...

This is one of the areas where I think the centroid ideas could come
in useful.  I don't completely understand the whole idea, but in
general, you have various index servers that reap data in a particular
area, including other index servers in related areas.  Then, from any
server that's close, you can zero in on the info you need.  They tell
you both what they know (or don't know) about your question and about
other servers that they think might be appropriate.

Let me try an example, pretty general just to make it smaller...I've
made the description worst-case for illustration.  For the example, my
question will be: "I'm looking for records with last name WHEELER" (I
actually am, but that's not important).  So I go to my favorite index
server and tell it that.  It's reply is "I don't have any last name
indexes, but here's a list of several that do."  I pick one and ask
it, and it's reply is "I have a bunch of name indexes, but not
WHEELER."  But since it's a name indexer it would know about more
servers and know more about them, so it could include a list of
index servers that have that name listed (and maybe indicate how many
entries or other properties, i.e "only Mass.")  And there I get a list
of al the Wheelers and can select any   I'm interested in.  This isn't
quite the "one-stop-shopping" of the ideal case, but it's only a few
steps (even in the worst case) and provides both scaling and
replication automatically.  It's also possible for this sequence to be
fully automated so the users perception is ask a single question, get
a final answer, all the succesive refinement happens under the covers.

   I propose a search protocol somewhat along the lines that Herb Stoyan is
   already implementing.

That's a query syntax, not actually a full search protocol.  As a
query syntax, I think it's too limited, but probably sufficient for
early testing.  In the end, I think we need to be able to specify a
wider range of fields, specifically including geographic areas,
because the number of matches could otherwise be far too high to deal
with.

   If this cannot be returned before the user's browser quits waiting,

I think if the design makes queries take that long it needs to be
fixed.  I don't expect this to be a problem, however.  End database
servers should have precomputed indexes if they have too much data to
build them on the fly in a timely manner.  Index servers should
certainly have this.

The important thing to start standardizing soon is something else.
For index services to work, there has to be a relatively uniform way
to get a listing of all records (in summary) that a site has.  That
way my server can collect this data from a bunch of servers and index
it and offer the overall indexing service.  Part of what's going to be
needed for this is uniform field naming and representation convention
(which GEDCOM does 99.44% of :-) and a syntax for the listing query.

To make this concrete, let me try out a general proposal...  Every
GenWeb server has a "base URL" which can be arbitrary, from there, the
path name syntax is defined as follows:

	-The first following component defines the type of access, it's
	chosen from a small set of keywords, the rest of the
	components are field specs dependant on the access type.  A
	server receiving a query with anb unknown access type should
	return a general information page on that server, and possibly
	including a pointer to some servers it thinks might be more
	extensive.

	-One access type is "record" which has one following
	component, the record ID.  This type accesses a specific
	record of the database, using whatever type of record ID that
	DB wants.

	-Another access type is "index" which asks the server to
	search it's index (either of just that server, or perhaps
	including other servers it indexes as well), the following
	fields specify the match criteria.

	-Another type is "list" which asks for a list of all matching
	records on the local server.  Following fields can specify
	selection criteria or what data is to be returned.

Some examples...  NOTE: as I made these up on the fly some of my "syntactic
sugar" choices may not really work.  Also, some of this data is real,
and some is fabricated.

http://map.db.genweb.org/genweb/record/1111
	-Would (in theory) get the record for me as I happen to be
	record 1111 in my DB

http://map.index.genweb.org/genweb/index/LN=Patton/B>1950
	-Would return a document like (links left out for readability
		in mail [except mine given for illustration]):
	
	The matching records in the local index are:
	

	
 Linda Patton [Bergman] (b: 1951)
	
 Richard (Ricky) Patton (b: 1954)
	
 Michael
			Anderson (Mike) Patton (b: 1954)
	
 John William (Jay) Patton (b: 1956)
	
 James David (Jamie) Patton (b: 1957)
	
 Karin Livingston Patton (b: 1958)
	
 Kevin Claud Patton (b: 1960)
	
 Stephanie Patton Flaum (b: 1977)
	
 Evan William Patton (b: 1987)
	
 Matthew Scott Patton (b: 1990)
	
	These matches came from
	Mike Patton's
	database, Andrew Smith's database and
	The Goff Gathering Association.
	

	There are additional Patton indexes on:
	

	
 Joe's server of Patton's in Rhode Island (20 entries)
	
 Ralph's  "comprehensive index of all Pattons (3000 entries)
	
	which may contain records not used in this index.

http://map.db.genweb.org/genweb/list/LN=Patton/B>1950/format:ID=NAME
	-Would return a text document like:
	1111=Michael Anderson (Mike) Patton
	1112=John William (Jay) Patton
	1113=Karin Livingston Patton
	1114=Kevin Claud Patton
	1121=Evan William Patton
	1122=Matthew Scott Patton
	4112=Richard (Ricky) Patton
	7108=Stephanie Patton Flaum
	7120=Linda Patton [Bergman]
	7122=James David (Jamie) Patton


Note the use of separate subdomains for actual databases and indexes
(in this case, they would probalby point to the same machine, which
would be www.map.cambridge.ma.us, when it's on the air :-).  Some
servers might only do one or the other, although I expect most will do
both.  That's a different topic I've mentioned before...

	-MAP

P.S.  I noticed that the root nameservers are delegating GenWeb.org to
UCSD, but the UCSD servers aren't configured to be servers for this
domain...



          Thursday, November 24, 1994 1:56:28 AM
          GenWeb Item
  From:           Herbert
Stoyan,hstoyan@faui80.informatik.uni-erlangen.de,Internet
  Subject:        indexing genweb
  To:             GenWeb
Mike Patton interesting proposals. We should develop this further and generate
a interface/protocol specification.

The proposal is to have the end of the URL: genweb/  -- result: html page of person?
index   -- Parameter LN=
                     B
           add:      FN=
                     D
                     M
                     C
           and other?             -- result: html-page with UL-list of 
                                             names, record-URLs, dates of
                                             *,+,oo, (as delivered in the call)
                                             sorted in year order
                                        + hints to other servers
list    -- Parameter as in index
            + format-description  -- result: text file of persons and record nos

Is the list case really necessary? Do we really need a specificable format
parameter? Shouldn't we restrict to some format?

As for the index-parameters: How to express burial? The international
genealogical symbols *,+,~,box,oo are not easy to write in parameter lists. How
to handle
blanks?

All this can be implemented quickly, withe one exception: the hints to other
indices. This requires knowledge about other indices at one site. How to
describe this?



          Thursday, November 24, 1994 3:13:53 AM
          GenWeb Item
  From:           Brian
Tompsett,B.C.Tompsett@computer-science.hull.ac.uk,Internet
  Subject:        Re: Indexing GenWeb Servers
  To:             GenWeb

 As has already been pointed out, any project with a centralised system is
flawed, by both offering a single point of failure and also be a victim
of its own sucess.

 There is also an assumption that indexing will be performed at the expence
of the data providers CPU time. This also not necessarily an appropriate
decision.

 There are merits to WAIS style indexing; where the search is performed at
the enquirers expense and not the providers.

 As a database provider I am wary of offering unlimted amounts of my server
for casual browsing.

  Brian



          Thursday, November 24, 1994 4:10:14 AM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        GENWEB.ORG domain (was: Indexing GenWeb Servers)
  To:             GenWeb
Mike Patton writes:
>P.S.  I noticed that the root nameservers are delegating GenWeb.org to
>UCSD, but the UCSD servers aren't configured to be servers for this
>domain...

I think we should solve this fairly quickly, as there are enough
lame delegations in the DNS already.  We don't need to fill the
domain with content right now, only set up a valid skeleton zone
somewhere.  Those with an interest in taking part in technical
DNS operations may want to have a look at

	http://www.docs.uu.se/~andersa/genweb/dns-op.html

where I try to maintain current information regarding the status
of the GENWEB.ORG domain.
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE


          Friday, November 25, 1994 12:58:05 PM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@ucsd.edu,Internet
  Subject:        Re: Indexing GenWeb Servers
  To:             GenWeb
Mike Patton wrote:
P.S.  I noticed that the root nameservers are delegating GenWeb.org to
UCSD, but the UCSD servers aren't configured to be servers for this
domain...

True. I am still working with the network managers here to add the
configurations necessary to serve this domain. Stay tuned.

Gary

***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************


          Friday, November 25, 1994 1:02:26 PM
          GenWeb Item
  From:           Gary Hoffman,ghoffman@ucsd.edu,Internet
  Subject:        Re:  David A. Chapin -- Automatic Reply
  To:             GenWeb
Anders Andersson wrote:

The Unix vacation(1) program should refrain from sending any
automatic responses to messages recieved due to membership in a
mailing list, but of course it can do so only if the message contains
some indication that it actually comes from a mailing list.  Maybe
some mail gateway along the road changes the To: line to contain
David Chapin's personal address rather than the normal genweb@UCSD.EDU
address?  I don't know.  This is not too important, but I think it's
a good idea to identify appearant problems and work towards a solution
before they grow too large and become a nuisance.


I reply:

The ucsd listserver, as opposed to other popular listservers, is homegrown
software and therefore behaves a bit differently than the others. One
difference is that mail sent to a ucsd mailing list appears to originate
from the originator rather than the list itself. This has the advantage of
allowing a reply to the sender, but the disadvantage of disguising the fact
that a given message may be a mass mailing rather than a personal message.
Also, vacation programs don't recognize a list message because the list is
not the sender.
Hope this explains the problem even if it doesn't cure the problem.

Gary


***************************************************************************
*Gary B. Hoffman, Computer/Language Lab Director e-mail: ghoffman@ucsd.edu*
*Graduate School of International Relations and Pacific Studies    (IR/PS)*
*University of California, San Diego (UCSD)          voice: (619) 534-7733*
*9500 Gilman Dr., La Jolla, CA 92093-0519 USA          fax: (619) 534-3939*
***************************************************************************


          Friday, November 25, 1994 2:50:24 PM
          GenWeb Item
  From:           Anders Andersson,andersa@Mizar.DoCS.UU.SE,Internet
  Subject:        Automatic vacation replies
  To:             GenWeb
[This is not really a GenWeb issue; sorry for the technical digression]

Gary Hoffman writes:
>The ucsd listserver, as opposed to other popular listservers, is homegrown
>software and therefore behaves a bit differently than the others. One
>difference is that mail sent to a ucsd mailing list appears to originate
>from the originator rather than the list itself. This has the advantage of
>allowing a reply to the sender, but the disadvantage of disguising the fact
>that a given message may be a mass mailing rather than a personal message.

Having the actual author rather than some pseudo-address in the From:
line is quite OK also for a mailing list.  No complaint here.

>Also, vacation programs don't recognize a list message because the list is
>not the sender.

There may be several different vacation programs around.  The one I
have in front of me is documented to recognize messages where the
individual recipient isn't mentioned in the To: or Cc: lines, as well
as those where the From: line includes the string "-REQUEST@", as
mailing list messages.  Either should be sufficient.

I'm not suggesting changing anything with the UCSD listserver.
I was talking to a postmaster elsewhere, and sent you a copy for
your information.  Again, this is not terribly important (although
it's probably a good idea to tell GenWeb contributors why they get
spurious vacation messages from people they haven't talked to).
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE



          Friday, November 25, 1994 3:44:37 PM
          GenWeb Item
  From:           Mike Patton€ genealogy mail,MAP=Genealogy@BBN.COM,Internet
  Subject:        Re: GENWEB.ORG domain
  To:             GenWeb
   Date: Thu, 24 Nov 94 13:09:09 +0100
   From: Anders Andersson 

   I think we should solve this fairly quickly, as there are enough
   lame delegations in the DNS already.  We don't need to fill the
   domain with content right now, only set up a valid skeleton zone
   somewhere.

Agreed, that's why I brought it up :-)...

   Those with an interest in taking part in technical DNS operations
   [Web page reference] where I try to maintain current information
   regarding the status of the GENWEB.ORG domain.

I looked at the page, looks like a proper summary of current status.
Perhaps Chris Garrigues could add a TEXT record in his skeleton DNS
pointing at that page.

Maybe we should get the people interested on a separate list.  I'm
sure many of the people on the GenWeb list (how big is this list,
anyway?) aren't really interested in the details of getting the Domain
going.  The list would also be a good entry for listing in the SOA for
contact.  Anyone care to host an MX for GenWeb.org for mailing lists?
I'd do this, but my home machines are only up on the net intermitantly
right now, and I h