Re(2): progress on HTML 2.0 reconstruction

James D Mason (MASONJD@oax.a1.ornl.gov)
Wed, 29 Mar 95 10:18:26 EST

I support Terry and Jon. If something is undefined in the agreed-upon
standard, then what it does to a browser is undefined. I'm looking forward to
the point at which browsers do better SGML parsing. It is true that ISO
8879:1986 does not require that parsers either validate or report errors, it
simply provides for recognition that some parsers might do so and gives them
credit for doing so. It is also true that there is no requirement that a
parser simply give up on finding an invalid document. But I would hope that a
good browser would at least report that it has received an unexpected tag (and
perhaps query the user for how it should be resolved).

Does this mean that Netscape is prohibited from messing around with odd tags?
Not at all. If they want to do something that only their clients can make
sense of, that's their option. But they need to weigh the consequences of
doing something that potentially makes documents unreadable to clients of
other vendors. That's their business decision, not one that conerns the HTML
standards community.

Having been in the standards business for a long time, I am always distressed
when someone proposes unilaterally breaking with the decisions made in a
public, open forum. Coming to a consensus about things for the good of us all
is, I thought, why we're all participating in this mailing list.

I think what this discussion suggests most is that vendors are ill advised to
hardcode their browsers to any single DTD. While we're obviously still in a
learning mode now and browsers are comparatively primitive, hard coding is
perhaps excusable. However, I think vendors would be better advised to make
browsers that can take more than one DTD (after the manner of SoftQuad's
Panorama).

I recognize, to be sure, the difficulties this brings with it, such as the
need to associate a document with a DTD and to associate a DTD with a
processing specification (such as a DSSSL specification). But I have watched,
as have Paul, Yuri, and others who participate in this discussion, what
happened in the CALS environment. In the early days, some vendors tied their
products to particular DTDs. Those DTDs soon changed, and then those vendors
were playing catch-up with the vendors who had accepted the spirit of the
standards and built open systems.

My point is that it would be a great shame if the WWW community failed to
learn from the experience of the CALS community.

To return to the intitial issue. I think we have no choice but to say that in
terms of HTML as a formally defined application of SGML that tags or other
consturcts not in the definition are "illegal". But we also need to accept
some flexibility in understanding what supporting the standard means.

r. 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