>The problem area came from trying to infer an SGML declaration
>from the MIME charset parameter (I think) as well as the numeric
>references problem.
Yes. I must apologise for not saying what I was thinking earlier. Most
of the discussion focused around wording the HTML 2.0 document such
that the SGML declaration inference mechanism became clear. I think
Francis and I went back and forth a little because I did not want any
commitment in HTML 2.0 until my proposal was aired.
>But if the document character set is ISO 8859-1, I'm not sure
>what the SGML interpretation is for a document in a different
>MIME charset, containing non-ISO-8859-1 characters.
The translation from the document character set to the system
character set is not specified by SGML. Hence, this is system
dependent.
>The other idea floated in the past for 2.0, of inferring an SGML
>declaration using the MIME charset as the document character set,
>would produce an interpretation (I think) of numeric character
>references not consistent with a later redefinition in 2.x
>of the document character set as Unicode, so we want (I think)
>to make sure we don't suggest that in 2.0 (even if it is
>current practice somewhere!).
Yes, I spent a lot of time trying to avoid anyone making a commitment
here!