Re(2): HTML 2.0 reconstruction done

James D Mason (MASONJD@oax.a1.ornl.gov)
Mon, 3 Apr 95 17:36:02 EDT

I wholeheartedly agree with the messages that warn against getting all bent
out of shape/divided into warring camps about SGML conformance.

Conformance to ISO 8879:1986 is not a dreadfully constraining thing, nor is
the operation of current browsers, however loose it may be, anything that
prevents HTML itself from being a conforming application of the ISO standard.

We need to remember that, Conformance, as defined in Clause 15, is rather
loosely defined. It is primarily conformance of documents and applications
(e.g. HTML), and only secondarily of software. This is intentional. In
document interchange it's most important that creators of documents send out
things in a predictable manner; that's why there's so much emphasis on things
like DTDs and SGML Declarations. How a particular piece of software goes about
interpreting the documents and their supporting declarations is another
matter, often out of the originator's control.

What the SGML standard has to say about software compliance is mostly for the
benefit of those who want to write error-reporting software, and it emphasizes
how they should state the capabilities of their software. DIS 13673, on
"Conformance Testing for SGML Systems", similarly is a reporting standard, not
an enforcement standard.

I think the current situation in the HTML application environment is in
compliance with the spirit of the SGML standard (and being in full compliance
with it should not be a burden). That is, most of what this WG has been
discussing has been the DTD and SGML Declaration. This WG is defining an
applcation, though with an eye to having that application interpreted by
software. SGML DTDs can be quite loose, as the current one is. That's not
necessarily an impediment to compliance with the standard.

Since the standard does not require that parsers and parser-based software
either validate or report errors, there's nothing to prevent a browser from,
for example, implementing tags that aren't in the DTD or not enforcing even so
little structure as is currently required. HTML's relationships with browser
software is more a matter of application semantics, and SGML isn't concerned
with that. The HTML standard is, as is appropriate, the documentation for the
semantics.

Although I am in complete agreement with Terry that I couldn't support an HTML
standard that said that the SGML standard could be ignored, I also believe
that what's being done in this WG is a long way from needing to ignore the ISO
Standard.

(BTW, I don't think there's any danger of a turf war between the ISO and the
ITEF in this area. In my ISO WG we _DON'T_ want to standardize applications.
DSSSL and HyTime, which are still pretty "meta", are about as close to
applications as we'll get. We believe that places like the ITEF are much
better for developing applications. The IETF understands its user community
and their requirements. We can't, as WG8, know that, and so we don't try to
standardize for it; we just build generic tools for others to use.)

So let's get off whether it's necessary to comply with/ignore ISO 8879. Let's
just make HTML as good an application as we can.

Jim

Dr. James D. Mason
(ISO/IEC JTC1/SC18/WG8 Convenor)
Oak Ridge National Laboratory
Information Management Services
Bldg. 2506, M.S. 6302, P.O. Box 2008
Oak Ridge, TN 37831-6302 U.S.A.
Telephone: +1 615 574-6973
Facsimile: + 1 615 574-6983
Network: masonjd @ ornl.gov