From: aray@nyc.pipeline.com (Arjun Ray)
Newsgroups: comp.text.sgml
Subject: Re: DTD for *incorrect* HTML
Date: 17 Feb 1996 00:30:30 -0500
Organization: FUDGE Dispersal Systems
Message-ID: <4g3p5m$rl7@alpha.nyc.pipeline.com>
References: \ <3123B24A.2BA0@passage.com> <4g0rsq$ncf@alpha.nyc.pipeline.com> <9602162308.AA08809@fly.HiWAAY.net>
[ posted and emailed ]
In article <9602162308.AA08809@fly.HiWAAY.net>,
Len Bullard \ writes:
>[Arjun Ray]
>
>>Travesty? No, just the true gauge of "support for open standards" and
>>sundry other feel-goodisms. And please don't ask how Netscape proposed
>>content providers should "hide" Javascript code from other browsers...
>
>We are actually puzzling over that one. Why was it considered a good idea
>to use an empty \' end-tag. Whatever. The contents of this container
are, in the best SGML-ese to express it that I know, a notation;
specifically, an object-oriented scripting language with C/C++-like
syntax. The issue, then, was how "best" to induce Javascript-challenged
processors to suppress this "in-lined" data when the default behavior
of HTML user-agents is to display anything that looks like, er, PCDATA.
At this point, a slight digression is in order. Most HTML user-agents
make a complete hash of comment declaration syntax (productions [91]
and [92].) The common (mis)understanding is to misconstrue a comment
declaration as a comment. Hence the notion of a "comment tag", whose
"STAGO" is `\'. A
marvelous example of this "syntax" is at \.
I recommend Netscape (1.1 or newer) and Lynx-FM. In Netscape, you'll
see what the Recent Features in Dr Dobb's Journal are. In Lynx-FM, you
won't. Time to "View Source". Relevant extract:
| \
\
| \Recent Features\
| [...]
| \
\
Yes, that's a single comment declaration with twenty-nine comments,
twenty-six of them null. However, in Netscape, a "comment tag" starts
with `\'... and a respected trade publication
seems to agree. And people learn from their betters, don't they?...
Back to the story. Some Bright Spark in Mountain View conceived the
Splendid Idea of using a "comment tag" to "hide" Javascript code. The
immediate problem, then, was to guard against something like
while(i-->0) { ... }
Thus, a Better Idea: precede the "comment close" with something
lexically distinct from legitimate Javascript. The proposal, as
published in Netscape's on-line documentation was this (please keep
the Maalox handy)
\
Later, the sentinel `\
\
\
#\
--
the Internet made me do it
From: churchyh@ccwf.cc.utexas.edu (Henry Churchyard)
Newsgroups: comp.text.sgml
Subject: Re: Non-SGML character found
Date: 17 Feb 1996 00:42:22 -0600
Organization: The University of Texas at Austin
Message-ID: <4g3tce$p03@piglet.cc.utexas.edu>
References: \
In article \,
Venant Habiyambere \ wrote:
>How to correct this error:
>sgmls: SGML error at /tmp/mifread.27352.sgml, line 113 at "\\014":
> Non-SGML character found; should have been character reference
>Is there any file where i can put the character reference.
Wouldn't it be best, if you want to allow form feeds in your
documents, to declare them as part of the character set in the
declaration? (I assume that's what "\\014" means.)
--
"Haughty Spain's fleet Advances to our shores, while ||churchyh@ccwf.cc.utexas.
England's fate, Like a clipp'd guinea, trembles in ||edu Henry Churchyard
the scale!" -- Sheridan "The Critic" (1779) http://uts.cc.utexas.edu/~churchh
Newsgroups: comp.text.sgml
Date: Sat, 17 Feb 1996 09:48:00 -0600
Message-ID: <9602171548.AA19965@fly.HiWAAY.net>
From: Len Bullard \
References: <4g2ff0$3f9@Venus.mcs.com> <31251A7D.7DDE@passage.com>
Subject: Re: Universal linktyping
[Eliot Kimber]
>Any attempt to define a "universal" set of semantics in any domain is
>always doomed to fail.
This is true Eliot, but consider that the reason for declaring a domain is
to limit the need for universality. We don't require a universal solution.
We require a set of semantics which are useful in some discourse. To limit
failure, we must know what a success is. ;-)
That the domain in an of itself is artificially constrained is assumed
(i.e., that might be why we say "artificial" intelligence, a domain we can
constrain by agreement, and "natural" intelligence for which we don't have
or nedd a good constraint. No rabbit trails here, I just want to point out
the issue you bring up in your next paragraph and emphasize the point of
having a testable/validatible agreement. Portability and interoperability
come down to an agreement about the domain and the range within in it
against which one can apply an operation one of which might be a test. Dan
Connolly makes this point clearly in a paper on character sets. I think it
is extendable to semantic sets.
>That's one of the reasons SGML is so cool: it's purpose is to enable the
>definition and use of new semantic constructs within the scope of
>reasonable constraints (i.e., the new constructs have to be formally
>declared and there is a stated expectation of human-understandable
>documentation of them somewhere (cf. definition 4.105, document (type)
>definition, in ISO 8879). This facility is extended and refined by
>enabling architectures by providing a facility for defining class
>hierarchies from which new semantic objects can be derived.
Yes. The important part of the enabling architecture is that it can be
used to provide a higher level of agreement. I mentioned in another post
the Netscape \ element type which has an associated \