Re: HTML 2.0 LAST CALL: Numeric character refs

Gavin Nicol (gtn@ebt.com)
Sat, 3 Jun 95 11:05:07 EDT

>No he hasn't. He showed that sgmls treats a non-conforming document
>differently than is suggested by the above verbage. That an HTML user
>may treat a non-conforming document differently than sgmls does not
>mean that the HTML spec is inconsistent with ISO8879.

I don't think it is non-conforming SGML (the standard is somewhat
vague). It is fine to say that HTML documents should be handled in such
a manner, but I think it goes against the intent of SGML, which is to
produce "correct" documents. The meaning of correct is subjective, but
I believe that in general, errors of any kind should be disallowed
(hence my earlier stance). As such, we must draw a line between
"correct", and the Internet way of thinking: be flexible in what you
accept. I think the best way is not to say anything, other than it is
an error, and leave it to browser implementors to decide upon the
behaviour. We had this discussion before, and there was no concensus
on how to handle out-of-range numeric character references, hence, we
should not pretend there is concensus now.