From: nikb@aladdin.co.uk (Nick barron)
Newsgroups: comp.text.sgml
Subject: Re: Recommendations wanted: editors for post-processing automated conversions
Date: Mon, 12 Feb 1996 07:34:22 GMT
Organization: SoNet - The first Internet provider on the south coast
Message-ID: <4fluio$c8t@news.aladdin.co.uk>
References: <4e3sfi$fsu@news.aladdin.co.uk> <310F87ED.1A19@spirit.com.au> <3114077B.22F7@passage.com>
"W. Eliot Kimber" \ wrote:
>I would agree that Author/Editor is well suited to this task (I depend on it).
Well, I was pushed to make a snap decision, and went for Author-Editor
-- looks like it`ll do the job nicely.
Many thanks for the advice.
From: asengupt@indiana.edu (Arijit Sengupta)
Newsgroups: comp.text.sgml
Subject: Re: nsgmls 1.0 and standard input
Date: 12 Feb 1996 00:53:24 GMT
Organization: Indiana University, Bloomington
Message-ID: <4fm324$kq7@usenet.ucs.indiana.edu>
References: \
Steve Heaney (heaney@cambridge.scr.slb.com) wrote:
: I've recently compiled and built the 1.0 version of nsgmls for Solaris 2.4
: and it works
: fine apart from one problem. When reading from standard input it
: consistently dumps
: core.
:
Use version 1.01 or get the patch from ftp.jclark.com or www.jclark.com
This was a known problem in 1.0. 1.01 fixes this.
Jit.
From: connolly@w3.org (Dan Connolly)
Newsgroups: comp.text.sgml,alt.hypertext
Subject: Re: Strongly-Typed Hyperlinks
Date: 12 Feb 1996 01:27:46 -0500
Organization: W3C
Sender: connolly@beach.w3.org
Message-ID: \
References: <31164A18.25C4@passage.com> <4f6onv$t18@Venus.mcs.com>
In-reply-to: jorn@MCS.COM's message of 5 Feb 1996 23:25:19 -0600
In article <4f6onv$t18@Venus.mcs.com> jorn@MCS.COM (Jorn Barger) writes:
> To make claims about the superiority of a hypertext
> design strategy without simulating it as far as possible
> on the WWWeb is ***scientifically irresponsible***
You don't even have to simulate it. Typed links have been a standard
architecture feature of the web since its inception. Popular
user agents just don't seem to exploit the feature.
A search for "typed link" in the www-talk archives:
http://www.eit.com/cgi-bin/wwwwais?keywords=typed+link\&selection=WWW-Talk%2C+1991-3\&maxhits=40
yields such goodies as:
http://www.eit.com/goodies/lists/www.lists/www-talk.1993q2/0282.html
HTML link "Made" relationship
Tim Berners-Lee (timbl@www3.cern.ch)
Tue, 11 May 93 16:46:44 +0100
>
> Can you cite *any* existing WWWeb site
> that would be more useful with typed links???
I experimented with typed links in producing the HTML renidition of the HTML 2.0 spec.
In that spec, evern glossed term is linked. It sure would be nice to have user
agents that could deemphasize those gloss links. Here's an example of the markup:
http://www.w3.org/pub/WWW/MarkUp/html-spec/html-spec_5.html#SEC5.7.3
\\SRC\
\
specifies the \URI\ of the image resource.
\(23)\
> And you allege that this will *add value*... but let's see some
> evidence.
Unfortunately, all I really have right now is an intuition that typed links
provide for richer knowledge capture and exchange, which is a Good Thing.
Dan
--
Daniel W. Connolly "We believe in the interconnectedness of all things"
Research Scientist, MIT/W3C PGP: EDF8 A8E4 F3BB 0F3C FD1B 7BE0 716C FF21
\ http://www.w3.org/pub/WWW/People/Connolly/
From: jon seymour \
Newsgroups: comp.text.sgml
Subject: SP 1.0.1 + Linux 1.2.13 + gcc 2.7.2 - segmentation fault
Date: Mon, 12 Feb 1996 20:56:20 +1100
Organization: The Warp Pharmacy
Message-ID: <311F0EC4.22D8EE63@zeta.org.au>
CC: jjc@jclark.com
Anyone have any experience with this combination?
nsgmls suffers a segmentation fault on all the tests that
run as a result of make check.
This is what I have installed:
linux-1.2.13 (ELF)
ld.so-1.7.14
gcc-2.7.2
libg++-2.7.1.3
libc-5.2.18
I did notice that the call stack gets corrupted by execution
of the declaration:
ParserToolkit kit;
in the first few lines of main(). Making kit a global fixed
the call stack corruption but did not stop nsgmls trapping.
Any ideas?
jon.
From: connolly@w3.org (Dan Connolly)
Newsgroups: comp.text.sgml
Subject: Re: Strongly-Typed Hyperlinks
Date: 12 Feb 1996 01:45:28 -0500
Organization: W3C
Sender: connolly@beach.w3.org
Message-ID: \
References: <4fdq2k$2vh@murphy.servtech.com>
In-reply-to: "Steven R. Newcomb"'s message of Thu, 8 Feb 1996 16:26:20 -0500
In article <4fdq2k$2vh@murphy.servtech.com> "Steven R. Newcomb" \ writes:
>
> Thanks for your reasonable, back-to-the-literature words. I think
> Randy Trigg's dissertation has not been as industrially appreciated as
> it should be. It was a prominent document in the HyTime committee's
> document register, and I still refer to it when I want to look at a
> list of garden-variety link types. The sheer open-endedness that one
> must confront when making such a list would, I think, always force an
> admission that there can be no "once-and-for-all" list of semantics to
> be expressed by linking.
It seems that there's some great experience in this area represented
by the readership of this froum.
A range of issues from standardized link types to compound document
architectures are being discussed in the HTML working group of the
IETF. There are two documents that appear to be on the standards
track, which means that the world may have to live with the decisions
that get recorded in those documents.
Please review (and send comments to html-wg@w3.org, after reviwing the
group charter etc. at: http://www.w3.org/pub/WWW/MarkUp/HTML-WG/):
( excerpted from http://www.ietf.cnri.reston.va.us/ids.by.wg/html.html
which, ironically, has a markup error that prevents three of the five
abstracts from being presented in my user agent :-)
ftp://ietf.cnri.reston.va.us/internet-drafts/draft-ietf-html-relrev-00.txt
"Hypertext links in HTML", M. Maloney, L. Quin, 01/15/1996. (64553 bytes)
The hypertext link mechanism is the connective tissue used to weave the
World Wide Web. A hypertext link is an object which specifies a
connection between any arbitrary addressable objects, locations, or
resources.
ftp://ietf.cnri.reston.va.us/internet-drafts/draft-ietf-html-cda-00.txt
"Compound Documents in HTML", P. Burchard, D. Raggett, 11/22/1995. (43493
bytes)
This specification provides an HTML implementation of a simple compound
...
Dan
--
Daniel W. Connolly "We believe in the interconnectedness of all things"
Research Scientist, MIT/W3C PGP: EDF8 A8E4 F3BB 0F3C FD1B 7BE0 716C FF21
\ http://www.w3.org/pub/WWW/People/Connolly/
From: mskuhn@unrza3.dialin.rrze.uni-erlangen.de (Markus Kuhn)
Newsgroups: comp.text.sgml
Subject: Re: ISO Public Text
Date: 12 Feb 1996 03:15:59 +0100
Organization: Markus Kuhn, 91080 Uttenreuth, Germany
Message-ID: <4fm7sv$m1@cortex.dialin.rrze.uni-erlangen.de>
References: <4flhov$ep5@reader2.ix.netcom.com>
Reply-To: mskuhn@cip.informatik.uni-erlangen.de
Carolyn Gordon \ writes:
>Where can I find ISO Public text entity declarations? Sorry
>for the dumb question, but I could not find a FAQ. Thanks in
>advance!
Erik Naggum's archive \ is a good place
to look for such files.
Markus
--
Markus Kuhn, Computer Science student -- University of Erlangen,
Internet Mail: \ - Germany
WWW Home: \
Newsgroups: comp.text.sgml,alt.hypertext
From: ht@cogsci.ed.ac.uk (Henry S. Thompson)
Subject: Re: SGML without DTDs?
In-Reply-To: Peter Murray-Rust's message of Sun, 11 Feb 96 10:18:27 GMT
Message-ID: \
Reply-To: ht@cogsci.ed.ac.uk
Organization: Centre for Cognitive Science, Edinburgh, UK
References: <9601300135.AA15133@fly.hiwaay.net> <4ek508$igd@Venus.mcs.com> <9601311617.AA22637@fly.hiwaay.net> <4eojl1$38s@Mercury.mcs.com> <9602010548.AA27472@fly.HiWAAY.net> \ <19960204T035437Z@arcana.naggum.no> <824033907snz@ursus.demon.co.uk>
Date: Mon, 12 Feb 1996 10:00:29 GMT
We have implemented and are making regular use of a system which is
very close to what you describe, which passes normalised SGML between
tools. See http://www.cogsci.ed.ac.uk/~ht/nsldoc.ps for an
introduction.
HOWEVER, we do use SP 'at the margins' of our architecture, and don't
expect people to produce normalised SGML by hand. Parsing
normalised SGML is fast and easy because we assume it's valid -- if it
isn't, we fall over hard. I'm skeptical about the possibilities of
constructing a robust parser for your normalised SGML which isn't
almost as complex as SP.
ht
--
Henry S. Thompson, Human Communication Research Centre, University of Edinburgh
2 Buccleuch Place, Edinburgh EH8 9LW, SCOTLAND -- (44) 131 650-4440
Fax: (44) 131 650-4587, e-mail: ht@cogsci.ed.ac.uk
URL: http://www.cogsci.ed.ac.uk/~ht/
From: bernd@NeRo.Uni-Bonn.DE (Bernd Kreimeier)
Newsgroups: comp.text.sgml
Subject: Re: DTD for *incorrect* HTML
Date: 12 Feb 1996 13:45:24 GMT
Organization: University of Bonn, Dept. of Comp. Sc. VI, Germany
Message-ID: \
References: \ <311E2EAB.14CC@passage.com>
In-reply-to: "W. Eliot Kimber"'s message of Sun, 11 Feb 1996 10:00:11 -0800
In article <311E2EAB.14CC@passage.com> "W. Eliot Kimber" \ writes:
> Joe Armstrong wrote:
> >
> > Is there a DTD for flawed HTML.
>
> \ \
> ]>
The w3c SGML declaration for HTML (3.0) says SUBDOC NO. Don't you
have to change this for your flawed_html.dtd as well as the w3c
HTML.DTD variations (strict, recommended)? To my understanding,
SUBDOC NO reflects the limitations of current HTML web browsers.
Perfectly complying, both Mosaic and current Netscape display minor
garbage at the document's start.
B.
From: bernd@NeRo.Uni-Bonn.DE (Bernd Kreimeier)
Newsgroups: comp.text.sgml
Subject: Re: SP 1.0.1 + Linux 1.2.13 + gcc 2.7.2 - segmentation fault
Date: 12 Feb 1996 13:49:58 GMT
Organization: University of Bonn, Dept. of Comp. Sc. VI, Germany
Message-ID: \
References: <311F0EC4.22D8EE63@zeta.org.au>
In-reply-to: jon seymour's message of Mon, 12 Feb 1996 20:56:20 +1100
This might be the same problem I ran into. Make sure to remove
any obsolete libiostream refs from the Makefile, and include
-lstdc++ instead. This fixed core's on
"nsgmls \" and all other problems
for me.
If there's a different problem, please let me know. I'd like
to put in the the "Known trouble" section of my
"SGML for Pedestrians" - if I ever find the time to write
the first release.
B.
From: milor001@maroon.tc.umn.edu (R A Milowski)
Newsgroups: comp.text.sgml
Subject: Re: nsgmls 1.0 and standard input
Date: 12 Feb 1996 07:57:21 -0600
Organization: University of Minnesota
Message-ID: <4fnh01$8sq@maroon.tc.umn.edu>
References: \
In article \,
Steve Heaney \ wrote:
>I've recently compiled and built the 1.0 version of nsgmls for Solaris 2.4
>and it works
>fine apart from one problem. When reading from standard input it
>consistently dumps
>core.
Yes, this is a bug. I believe there is a 1.1 version that fixes this.
--
==============================================================================
R. Alexander Milowski milor001@maroon.tc.umn.edu
Copernican Solutions Incorporated (612) 825 - 4132
From: mcclellantj@harrier (Tad McClellan)
Newsgroups: comp.text.sgml
Subject: Re: What are the limits of marked section declarations?
Date: 12 Feb 1996 14:19:33 GMT
Organization: Lockheed Martin Tactical Aircraft Systems
Message-ID: <4fni9l$imh@cliffy.lfwc.lockheed.com>
References: <4fg9pb$11h@news.cerf.net>
Reply-To: mcclellantj@lfwc.lockheed.com
Whatis (whatis@yyz.com) wrote:
: I'm currently converting a help system with a proprietary source
: into SGML. Most of it has gone pretty smoothly, but I'm up
: against a problem I can't solve.
: The previous help system allowed the help to be tuned to specific
: products and families of products, using boolean expressions at
: about the same level as the C pre-processor. For instance,
We, in the DoD IETM arena, refer to this as 'context filtering'. We
use many different contexts such as: aircraft effectivity, technician
skill level, whether the last test passed or failed, etc...
: @X||Y&&(!Z)
: Some text
: @ELSE
: Some other text
: @END
: The facility I see in SGML for doing stuff like this is the marked
: setion declaration, combined with parameter entities. That way, I
: can set up declarations like
: What do I do in SGML? Or am I SOL?
As another followup pointed out, you're maybe just FOL (Fart OL),
not quite as serious as SOL...
The Content Data Model (CDM) DTD (MIL-D-87269) puts the burden of
evaluating the expressions on the SGML application (Presentation
system, in our case), rather than on the SGML parser.
These are primarily implemented with the \, \,
\ and \ elements.
e.g.
\
\
\
\
\
\
\
\NUM_SEATS\
\
\1\
\
\
\
The Aircraft has one control stick.
\
\
\
\
\
\NUM_SEATS\
\
\2\
\
\
\
The Aircraft has two control sticks.
\
\
\
\
\
\
\ is a variable name that the SGML application has set
previously (as initialization, or by another IETM document).
The SGML app evaluates the \ and if TRUE presents the \
else skips \
Kinda ugly, but workable...
--
Tad McClellan, Logistics Specialist (IETMs and SGML guy)
email: mcclellantj@lfwc.lockheed.com
The "Battle of the Sexes" is perpetuated by fraternizing with the enemy.
From: stuart@telerama.lm.com (R. Edward Stuart)
Newsgroups: comp.text.sgml
Subject: Re: Musing on Networked Hypertext
Date: Mon, 12 Feb 1996 03:49:40 GMT
Organization: Telerama Public Access Internet, Pittsburgh, PA
Message-ID: <4fnk19$66@scramble.lm.com>
References: <31014D2A.7BF2@passage.com> <9601220123.AA21903@fly.HiWAAY.net>
>"The word is not the thing" -- S.I. Hayakawa.
So who the fuck is "Hayakawa?"
Thanx,
Ed Stuart
stuart@telerama.lm.com
From: stuart@telerama.lm.com (R. Edward Stuart)
Newsgroups: comp.text.sgml
Subject: Re: What's New in the SGML World?
Date: Mon, 12 Feb 1996 03:49:48 GMT
Organization: Telerama Public Access Internet, Pittsburgh, PA
Message-ID: <4fnk1h$66@scramble.lm.com>
References: <5SviKAAP8kBxEwzL@kglps.demon.co.uk> <4ejgdb$43m@franklin.its.utas.edu.au>
3. What publications would you recommend?
>"Practical SGML"
Thanx,
Ed Stuart
stuart@telerama.lm.com
From: mcclellantj@harrier (Tad McClellan)
Newsgroups: alt.hypertext,comp.text.sgml
Subject: Re: REQ: FAQ
Followup-To: alt.hypertext,comp.text.sgml
Date: 12 Feb 1996 14:44:10 GMT
Organization: Lockheed Martin Tactical Aircraft Systems
Message-ID: <4fnjnq$46l@cliffy.lfwc.lockheed.com>
References: <311D39E7.590F@hobbe.adb.gu.se>
Reply-To: mcclellantj@lfwc.lockheed.com
Jonas Liljegren (a95jonas@hobbe.adb.gu.se) wrote:
: Is there a FAQ for this newsgroup?
Not an official one. There is a wealth of info/links at:
http://www.sil.org
--
Tad McClellan, Logistics Specialist (IETMs and SGML guy)
Birthdays are good for you ...
A federally funded study has shown that people with the
most birthdays live the longest.
From: mcclellantj@harrier (Tad McClellan)
Newsgroups: comp.text.sgml,alt.hypertext
Subject: Re: SGML without DTDs?
Followup-To: comp.text.sgml,alt.hypertext
Date: 12 Feb 1996 15:34:00 GMT
Organization: Lockheed Martin Tactical Aircraft Systems
Message-ID: <4fnml8$46l@cliffy.lfwc.lockheed.com>
References: <9601300135.AA15133@fly.hiwaay.net> <4ek508$igd@Venus.mcs.com> <9601311617.AA22637@fly.hiwaay.net> <4eojl1$38s@Mercury.mcs.com> <9602010548.AA27472@fly.HiWAAY.net> \ <19960204T035437Z@arcana.naggum.no> <824033907snz@ursus.demon.co.uk>
Reply-To: mcclellantj@lfwc.lockheed.com
Peter Murray-Rust (Peter@ursus.demon.co.uk) wrote:
: I am developing SGML applications for the molecular sciences (a set of
: DTDs I shall generically call Chemical Markup Language (CML)). I have come
: to a stage where I wish to have the system implemented as a means for
: transferring molecular data between many different applications on many
: different platforms. I can assume that my community is computer-literate,
: understands the need for absolute precision in constructing documents,
: is (for the most part) well disposed to the implementation of standards,
: and is _almost totally_ ignorant of SGML. (It is however, thoroughly
: acquainted with HTML and will understand the general concept of hierarchical
: data structures with balanced delimiters).
: PROBLEM:
: I have a major problem in that to distribute this requires a great deal of
: complex and error-prone SGML technology which I do not think is necessary
: for my application. (The errors come not from SGML itself, but its use by
: a community who won't know the first thing about it :-).
: QUESTION:
: Is it possible to agree a 'subset' of SGML that will allow this community
: (and I suspect many others) to use the SGML approach with simpler technology?
: PROPOSAL:
: I would like to agree on a subset of SGML that was parsable into an ESIS
: stream without the use of DTDs. This subset would still be valid SGML and
: would parse against its DTD(s) using SGML. It relies on documents being
: valid normalised SGML without default attributes. It also assumes that there
: is a relatively limited use of entities and, I suspect, no marked sections.
: In making this proposal I am probably repeating existing suggestions and I'm
: also likely to have overlooked aspects of SGML that may cause problems, so
: please forgive these in advance.
: ELEMENTs:
: All elements should have explicit end-tags, though these could be abbreviated
: to withoutproblems.
Or even abbreviated further if pre-processed with a normalizer such
as James Clark's 'spam'.
: However, I have a problem with EMPTY content models.
^^^^^^^^^^^^^^^^^^^^
\
Technically, EMPTY is 'declared content', not a 'content model'. See
ISO-8879 production [116] and [125]
\
: sgmls gives an error for including an end-tag for these (I assume this
: requirement is in the standard). If a parser does not have a DTD it has to
: look ahead to the end of the document before it can be sure that there is
: no closing tag for an element.
You have a similiar problem (not known if balance required/disallowed)
with #CONREF attributes.
: ATTLIST
: The primary use of ATTLIST that I make is to (a) validate that the
: attribute name is correct, and (b) to supply default values. If these
: are guaranteed by the creating software there is no need for ATTLIST.
: (SGML has weak type-checking so the loss of that is unimportant).
Another use of ATTLIST is to identify #CONREF attributes, as above.
Moot, if your DTD uses no CONREFs.
: Conclusion:
: -----------
: This proposal might be though to have removed most of the valuable bits of
: SGML, but what is left is extremely valuable for our community. Until
: I started using SGML and CoST I was unaware of how valuable it was to have
: structured hierarchical documents. With the use of CoST/DSSSL I can now
: ask much more powerful queries of primary data that I could with (say)
: the RDB model (mainly because it is such an effort to build RDBs). Moreover,
: SGML solves the problem of separating syntax from semantics which is still
: unresolved in many other systems.
--
Tad McClellan, Logistics Specialist (IETMs and SGML guy)
email: mcclellantj@lfwc.lockheed.com
The "Battle of the Sexes" is perpetuated by fraternizing with the enemy.
Newsgroups: comp.text.sgml
Date: Mon, 12 Feb 1996 11:24:36 -0600
Message-ID: <9602121724.AA22318@fly.HiWAAY.net>
From: Len Bullard \
References: <4fdq2k$2vh@murphy.servtech.com> \
Subject: Re: Strongly-Typed Hyperlinks
[Dan Conolly]
>A range of issues from standardized link types to compound document
>architectures are being discussed in the HTML working group of the
>IETF. There are two documents that appear to be on the standards track,
>which means that the world may have to live with the decisions that get
>recorded in those documents.
Thanks eversomuch for the ftp addresses to the cited documents. We can
look at these and compare them to requirements for HyTime-conforming
systems. However, just one quibble: the WWW, not the world or even the
Internet, lives with the decisions of the HTML working group. These are
separable domains.
Now ask if it is the current biggest game in town... ;-)
Len Bullard
From: jorn@MCS.COM (Jorn Barger)
Newsgroups: comp.text.sgml,alt.hypertext
Subject: Re: Strongly-Typed Hyperlinks
Date: 12 Feb 1996 10:36:21 -0600
Organization: The Responsible Party
Message-ID: <4fnqa5$p8t@Venus.mcs.com>
References: <31164A18.25C4@passage.com> <4f6onv$t18@Venus.mcs.com> \
In article \,
Dan Connolly \ wrote, quoting me:
> > To make claims about the superiority of a hypertext
> > design strategy without simulating it as far as possible
> > on the WWWeb is ***scientifically irresponsible***
>You don't even have to simulate it. Typed links have been a standard
>architecture feature of the web since its inception. Popular
>user agents just don't seem to exploit the feature.
That's reason enough to simulate, isn't it?
[...]
>I experimented with typed links in producing the HTML
rendition of the HTML 2.0 spec. >In that spec, every glossed term is
linked. It sure would be nice to have user >agents that could
deemphasize those gloss links. [...]
>http://www.w3.org/pub/WWW/MarkUp/html-spec/html-spec_5.html#SEC5.7.3
Interesting... let me think out loud about this:
So, browsers could have a 'glossary mode' that reveals which terms are
defined... or you could configure a non-obtrusive style for them, like
dark-blue text...
It sounds like a lot of markup, unless you can assign the HREF in the
header... And even then, for a word that's used 1000 times on a
page, it sounds like a pretty inefficient approach. How about an
embedded glossary-object that tells which words are glossed (and
where)?
You can learn a lot about the usefulness of this by simulating it with
regular links as you've done (except they're a lot more obtrusive) or
with an asterisk or something next to defined words, and the glossary
at the bottom of the page...
My sense, though, is that this isn't really a problem you want to try
to 'mechanize'...
I think there's a general danger in hypertext theory of trying to take
the burden off the writer, by creating 'style-templates' you can just
*fill in*.
But this *encourages* atrocious writing habits! Writers are supposed
to think carefully about who's going to be reading their work, and
fine-tune it to meet their needs.
How many people are even going to bother to create the glossary links?
Wouldn't it work better to just have a convention that all documents
should include a link labelled "If you don't understand the
vocabulary, try this first"?
What other sorts of semantic links are being proposed?
[...]
> > And you allege that this will *add value*... but let's see some
> > evidence.
>Unfortunately, all I really have right now is an intuition that typed links
>provide for richer knowledge capture and exchange, which is a Good Thing.
Well, I appreciate your honesty!
(So, as far as you know, the people who claimed that the WWWeb was
fatally flawed for lack of them, and beneath the interest of the
experts, were talking Bull? ;^/
j
-==---
. hypertext theory : artificial intelligence : finnegans wake . _+m"m+_"+_
lynx http://www.mcs.net/~jorn/ ! Jp Jp qh qh
best-of news:alt.music.category-freak ! O O O O
ftp://ftp.mcs.com/mcsnet.users/jorn/ Yb Yb dY dY
...do you ever feel your mind has started to erode? "Y_ "Y5m2Y" " no.
Newsgroups: comp.text.sgml,alt.hypertext
From: jbottoms@world.std.com (John W Bottoms)
Subject: Re: SGML without DTDs?
Message-ID: \
Organization: The World Public Access UNIX, Brookline, MA
References: <9601300135.AA15133@fly.hiwaay.net> \ <19960204T035437Z@arcana.naggum.no> <824033907snz@ursus.demon.co.uk>
Date: Mon, 12 Feb 1996 16:35:49 GMT
In article <824033907snz@ursus.demon.co.uk>,
Peter Murray-Rust \ wrote:
>I am developing SGML applications for the molecular sciences (a set of
>DTDs I shall generically call Chemical Markup Language (CML)). I have come
>PROBLEM:
>The major problem I have is that to run SGML properly every site has to
>implement:
> a parser
> DTDs for every document type that I use
> a catalog (to keep the whole thing together)
> a set of *.ent files (e.g. for ISOlat1, etc)
>QUESTION:
>Is it possible to agree a 'subset' of SGML that will allow this community
>(and I suspect many others) to use the SGML approach with simpler technology?
>PROPOSAL:
>I would like to agree on a subset of SGML that was parsable into an ESIS
>stream without the use of DTDs. This subset would still be valid SGML and
>would parse against its DTD(s) using SGML. It relies on documents being
>valid normalised SGML without default attributes. It also assumes that there
>This approach allows 'unknown' GIs, and it may be claimed that this
>completely breaks the spirit of SGML. Since the files would have to parse
>Peter Murray-Rust, GlaxoWellcome, domestic net connexion
(snipped heavily)
It is not clear whether you are talking about using a DTD or not. But
anyway...
1. You have a mixed model here. SGML is parsed. The instance can be
translated into another data stream. I'm not familiar with ESIS but
I'll assume that there is a mapping from some instance (against some
DTD) that can be mapped into an ESIS form.
2. All SGML instances require a DTD. So what you are proposing is a
modification to SGML or another metagrammar...not SGML. I think
what you are really talking about is an application that works
with a non-SGML markup that will do what you need. We could refer
to the path HTML has taken for this one. Maybe the real question
should be "Why bother with SGML at all?". It may be more baggage
than you need. And I think you have a DTD in mind but it may be
buried in your proposed application.
Finally, it sounds as though you would like to be able to use tags
which can be anything that is of a legal format and length. SGML
does not allow regular grammars for the expression of elements so
this would not be possible in SGML. Again, the driving force here is
to look at an alternate metagrammar or explain why you would like to
use SGML. Some more detail about the implemented system would be
helpful.
-jb SGML'er
From: chip@ils.nwu.edu (Chip Cleary)
Newsgroups: comp.text.sgml,alt.hypertext
Subject: Re: Strongly-Typed Hyperlinks
Date: Mon, 12 Feb 1996 11:32:57 -0600
Organization: The Institute for the Learning Sciences
Message-ID: \
References: <31164A18.25C4@passage.com> <4f6onv$t18@Venus.mcs.com> \ \ \
In article \,
Bernstein@eastgate.com (Mark Bernstein) wrote:
> SEPIA: Streitz, Thuering, Haake, Haake, et al. created an important
> collaborative hypertext system which used "activity spaces" to organize
> concurrent collaboration, and relied on link types to clarify to everyone
> (and to the system) the purpose and role of each hypertextual component.
> See especially "What's Eliza Doing in the Chinese Room" in Hypertext 89
> (or 91?)
For an example of a WWW system that uses link types in a similar way (to
clarify where the user can look for different types of information), see:
http://www.ils.nwu.edu/~e_for_e
This system uses link types to collect specific follow-up questions about
a page under general question categories. This organization is intended
to:
- allow users who want to ask specific questions to quickly locate them, and
- to provide a set of general types of questions that less directed
users can scan to see which, if any, they want to ask (and then to
show under these general questions the specific instantiations of them
that the system can actually answer).
This system is a quickie WWW implementation of a modified type of ASK
system [Bareiss & Osgood, Hypertext-93].
In article \,
Bernstein@eastgate.com (Mark Bernstein) wrote:
> Not within the constraints of http. The web is carefully engineered to
> avoid the need to keep any state information between transactions. This
> means that web servers need not maintain account information or session
> records for each user, which makes web servers *much* faster and more
> efficient than, conventional session-based designs.
Although web sites *need not* maintain state information, I think they
*can* if you want them to. For example, ACM's "living catalog" tracks
users by generating web pages that have a user id added to every lin. It
uses this capability to track one type of state -- letting users carry
around a "basket" of things they've ordered as they browse.
http://www.acm.org/catalog/
A site that wanted to customize pages for individuals based on what other
pages they've already seen could use a similar approach (as long as the
user remains within the confines of the site or logs back in). I've wanted
to add this type of capability to e_for_e for some time to better suggest
how individual users may move "forward" through the system, though I've
never found the time. Right now, each page just has a hard-coded "next"
page.
- Chip
Newsgroups: comp.text.sgml,alt.hypertext
From: mlt@netcom.com (Marcy Thompson)
Subject: Re: Strongly-Typed Hyperlinks
Message-ID: \
Organization: Netcom Online Communications Services (408-241-9760 login: guest)
References: <31164A18.25C4@passage.com> <4f6onv$t18@Venus.mcs.com> \
Date: Mon, 12 Feb 1996 18:46:38 GMT
Sender: mlt@netcom8.netcom.com
Jorn Barger writes:
> To make claims about the superiority of a hypertext
> design strategy without simulating it as far as possible
> on the WWWeb is ***scientifically irresponsible***
Is this like saying: "To make claims about the usefulness of some
new user interface feature for word processors without simulating
it as far as possible using Word macros is scientifically irresponsible"?
What if the new user interface feature were incompatible with Word as
it is currently available? What if I knew and could prove that any
test on Word would be so seriously distorted as to be meaningless or
worse?
Note that I have said nothing about whether I think that the WWW
is a useful test bed for a particular theory of hypertext design. I
think it *is* a useful test bed for some of them. But I do not believe
that it is a uniformly useful test bed for all theories of hypertext
design, and *I* think that claims that it is are irresponsible in
the extreme.
Test beds vary in their effectiveness. To know whether the WWW is
a useful test bed for a particular theory, you have to know things like
whether the theory is compatible with the design of the WWW, whether
the theory is testable on the WWW, what aspect of the theory you need
to test, and so on.
Marcy (who is *not* Eliot, okay, so would people please stop
sending me mail about it?)
--
Marcy Thompson
work: marcy@passage.com
play: marcy@squirrel.com mlt@netcom.com
From: Peter Murray-Rust \
Newsgroups: comp.text.sgml
Subject: Re: SGML without DTDs?
Date: Mon, 12 Feb 96 18:56:54 GMT
Organization: Myorganisation
Message-ID: <824151414snz@ursus.demon.co.uk>
References: <9601300135.AA15133@fly.hiwaay.net> \ <19960204T035437Z@arcana.naggum.no> <824033907snz@ursus.demon.co.uk> \
Reply-To: Peter@ursus.demon.co.uk
[First an apology; my mailer included alt.hypertext in this thread so
please remove it in any replies]
Thank you for your replies so far. I have had a chance to peruse
NSGML (Henry Thompson) which is very much along the lines of what I
wish, although I think I can make my requirements simpler. I reiterate
that all documents are valid SGML and parse using SGMLS (I am not
inventing another language). What I wish to do is:
define a valid _restricted subset_ of SGML, used to create DTDs.
create document instances which parse against these DTDs and
produce an ESIS stream (e.g. using SGMLS)
write alternative (simple) software that has no knowledge
of the DTD, but produces the same ESIS stream.
To reformulate the question with an example of a document instance:
\
\
]>
\
\xyzzy\
\plugh\
\
This parses with sgmls.1.1.91 and produces the ESIS stream:
(FOO
(BAR
-xyzzy
)BAR
(BAR
-plugh
)BAR
)FOO
which I can read into CoST and translate/render/search, etc.
Note that (so far) the \...\ cannot be unambiguously translated
into an ESIS stream without the knowledge of the DTD. For example, a
DTD like:
\
\
\
will produce a difference ESIS stream (adding additional whitespace content
and the default attribute ATT="NO" for both BAR's).
However, if we forbid mixed content (#PCDATA|BAR), and default attributes,
this DTD is not valid (under our rules). The question is whether it is
possible to create a set of rules whereby the ESIS stream can always be
generated from the document instance.
(Note that the ESIS stream does not retain all the information in the
document instance, so the reverse process (ESIS->*.sgm) is not normally
possible. With a restricted set of DTDs it might be possible to recreate
useful *.sgm files).
NSGML defines 'semi-valid' SGML. The DOCTYPE is omitted, and there can
be multiple instances of subtrees (e.g. there could be several \...\
in an instance). This is not allowed in what I am proposing - there must
always be a valid DOCTYPE and the document must validate against SGMLS.
(The non-SGMLS translation software will ignore the DOCTYPE).
Bringing together the rules in NSGML, with the additional omission of
#CONREF (Tad McClellan) and some of my own, the following set of rules
is proposed as a starting point.
- Document is coded using an ISOlatin character set, with embedded
character entities where necessary
- Reference concrete syntax is used
- No capacity or length restrictions
- All NAMES, etc must be valid SGML
- No short refs (?or tag minimisation)
- No SUBDOCs
- No marked sections
- All end-tags present (except for EMPTY elements)
- All entity references terminated with ';'
- No entity references except of type SDATA (i.e. everything is in
a single file except agreed character sets)
- All NAMEs (start and end tag GIs, attribute names) have no lower
case letters
- No CDATA or RCDATA element content
- NO #CONREF
- No mixed content (#PCDATA|FOO)
- No default attribute values
- All attribute values to have CDATA type
- All potential SGML delimiters (e.g. '<') to be escaped (e.g. \<)
Although this restricts the possible things that could be done with SGML
I believe that DTDs (and document instances) constructed in this way would
be powerful enough to bring real benefits.
This is not meant to undermine the validation procedure. A typical way that
it would be used is for an international supplier of data to provide
fully validated document instances for distribution over the WWW. These
would be flagged as readable without SGMLS and it would be possible to create
simple software which created the ESIS stream. Parsing this (e.g. with CoST)
for rendering/translation/searching, etc. would then be identical to
postprocessing SGMLS output.
Peter.
--
Peter Murray-Rust, GlaxoWellcome, domestic net connexion
From: Bernstein@eastgate.com (Mark Bernstein)
Newsgroups: comp.text.sgml,alt.hypertext
Subject: Re: Strongly-Typed Hyperlinks
Date: Mon, 12 Feb 1996 14:11:18 -0500
Organization: Eastgate Systems, Inc.
Message-ID: \
References: <31164A18.25C4@passage.com> <4f6onv$t18@Venus.mcs.com> \ <4fnqa5$p8t@Venus.mcs.com>
Ok, folks, I've just about had it.
On the subject of strongly-typed links, Dan Connolly writes:
> >Unfortunately, all I really have right now is an intuition that typed links
> >provide for richer knowledge capture and exchange
Jorn Barger "paraphrases" this response as follows:
> So, as far as you know, the people who claimed that the WWWeb was
> fatally flawed for lack of them, and beneath the interest of the
> experts, were talking Bull?
First, Barger implies that some believe that WWW is "beneath the interests
of the experts". He says this often. It is false, silly, and deceptive,
but that seems no deterrent.
Disposing of the gratuitous clauses, Barger writes that
> So, as far as you know, the people who claimed that the WWWeb was
> fatally flawed for lack of [typed links]..... were talking Bull?
This is reprehensible nettiquette, as it purports to paraphrase Connolly's
position while actually contradicting it.
This is irresponsible argumentation. How does "talking bull" contribute to
the discussion?
This is mendacious, since it sets up a preposterous straw man (the
mythical hypertext expert with whom Barger seems utterly obsessed) to
distract people from the indefensible underlying claim: that Barger has
added anything to the extensive research record on typed links, or indeed
to what he terms "hypertext theory".
--
Mark Bernstein Bernstein@eastgate.com
Eastgate Systems, Inc voice: (800) 562-1638 +1 (617) 924-9044
134 Main Street fax: (617) 924-9051
Watertown MA 02172 USA http://www.eastgate.com/
Newsgroups: comp.text.sgml
Date: Mon, 12 Feb 1996 15:17:35 -0600
Message-ID: <9602122117.AA15266@fly.HiWAAY.net>
From: Len Bullard \
Subject: Re: Facts, folks. We need facts.
Cross-posted to comp-text-sgml for those in the SGML community who still
question why DTDs are needed. They allow communities to express and make
validatible agreements. At will...
The VRML community is now witnessing what cowboy standards achieve: they
round up the cows, pen them, then ship them off to the meat-packing plant.
[Jan H.]
>I have talked to someone who claims to have seen a press release from
>Netscape confirming every rumor ever whispered (except for the Snapple
>deal).
>
>If anyone has this, they should post it, or a URL. I'll keep looking.
Well, I have read the SGI posting on the subject.
This is process? They panicked, or as the old jazz standard says, "You
made your move too soon."
The sad thing is that Moving Worlds would probably have won honestly. Now,
there can be no honest decision. It was a grand try though. Some folks
have no faith in consensus, and they were risking too much investment. If
the 4 billion dollar bubble bursts, some programmers may have to go back to
U of ChiTown and get their jobs back. Poor old WWW. It makes a nice
hostage.
Apple and Microsoft should sue, but they won't. In a situation with no
legal entity, there is no recourse to due process. Understand that this is
why real standards bodies are provenanced. What M\&A can do is walk away
from a game rigged for them to lose. They can make a deal to combine their
proposals and field a new 3D metafile format for the Internet. The
advantages of such a deal are obvious. A company in trouble that coalesces
with its stronger rival and in that combination, beats their mutual
competitors would be doing smart business even if only for a few deals.
You don't have to merge to dance.
You do have to own the framework API.
The Internet is not the WWW. Some state that VRML is for the "Net".
Smarter folks can figure out that there are many applications for 3D
metafiles that don't need, and in fact are hindered by, network engines.
There are better models for hyperlinks, there are better protocols than
HTTP, and there are better applications than HTML and VRML. The W3C and
its 120 consortium members want you to believe that success comes from
consensus.
Success comes from winning market share.
Cooperation is a way to get that. So is competition. Competition has an
advantage: competitors evolve, adapt, and take and hold resources that they
deny to other competitors. It isn't necessary to eliminate a competitor,
just starve them until they become cooperative.
Tools. The secret to success is the tools.
Novices don't know that, but they understand ease. In hypermedia, the
overwhelming number of customers are novices. By the time they figure out
they need more, they have invested too much. Remember all those "cool
people" in the seventies and eighties who swore to you that cocaine wasn't
addictive?
Smart folks know otherwise.
The content providers could choose to walk away, but we probably won't. Of
course, if someone else provides a complete tool suite at a decent price
and a complete environment guaranteed for the next release of their
operating systems, we might be persuaded otherwise. We are a fickle lot.
Making the lips of the fish move can't be that hard, and I doubt that the
current amount of extant content is that large. I'd rather develop content
for a corporation with depth and breadth in the product line than for one
banking on a browser.
As for the gods, we are men. As for heroes, beware. There was a fellow
named Alcibiades who comes to mind just now.
I vote for ActiveVRML. Call your friends and ask them to do the same. At
least all the betas will come in one box.
Len Bullard
Newsgroups: comp.text.sgml
From: donald@sq.com (Donald Teed)
Subject: Re: Tanks: TEI & SQ's Author/Editor
Message-ID: <1996Feb12.210415.23133@sq.com>
Organization: SoftQuad Inc., Toronto, Canada
References: \ \
Date: Mon, 12 Feb 1996 21:04:15 GMT
ST.KURTZ@BIONIC.zerberus.de (Stefan Kurtz) writes:
>A big "thank you" to all the people (especially to the SoftQuad people and to Arjan
>Loeffen), who sent me mails with solutions and hints.
>Well, Author/Editor _works_ with TEI.
I'm glad to hear this came to be. Just so that others in the TEI
user community can learn about the problem and the solution, we thought
it would be good to share the news on this issue with others.
The TEI DTDs are a set of DTDs that use undefined elements as well
as marked section parameter entities in the document type declaration
subset. These usages are not presently supported by SoftQuad's RulesBuilder,
although we are working on supporting these. (Marked section
parameter entities that are applied on the document instance are supported,
but those which would have effect within the DTD are not supported.)
Thus the TEI DTDs will not compile (as is) with RulesBuilder.
The work-around is:
1) do not use parameter entities in the document type declaration subset
2) do not use elements without defining a content for them.
One solution is to use a preprocessor such as Greg Murphy's preparse,
a Perl script available from ftp://ceth.princeton.edu/tei/tools/preparse,
which expands all marked sections and external parameter entities, and cleans
the content models of all elements of any undefined elements.
The reason that the mkrls program does not appear to have a problem with
the undefined elements is that some versions of this program had a bug
with a command line option to suppress warnings about defined
and unused elements. The bug caused this same switch to
suppress errors about undefined elements. Some people in
the TEI community have reinterpreted this bug as a feature.
From our point of view, RulesBuilder 3.0 fixed the bug with
the "Report elements declared but not used" feature.
I'd like anyone having problems with supported SoftQuad software
to make use of our product support. We can be contacted at
support@sq.com
--Donald Teed, Technical Support, SoftQuad