Re: WD-xml-infoset-20010316

> Should not unparsed entity and notation information items have a [base
> URI] property, which an application would use for resolving a relative
> system identifier? Also unexpanded entity reference info items probably
> need a base URI as well (corresponding to the base URI of the entity
> declaration).

Good point.  At one time they had a base URI which was meant to
be the base URI of the entity itself, but that can only be determined
when the entity is fetched, because of HTTP redirection (and possibly
catalogue translation).  Perhaps it should be called something like
[declaration base URI] to emphasize that it is not the base URI of
the entity itself.

> The note in 2.6 that end-of-line normalization is applied to character
> info items applies equally to the content of comments and processing
> instructions.  Maybe this is worth mentioning in the spec, since it's a
> detail that implementors can easily forget.

Yes, and maybe we should move it to the introduction.

> The [character encoding scheme] and [system identifier] properties on
> document information items trouble me a little.  Logically they are
> associated with the document entity not with the document.  I think the
> right way to deal with the removal of parsed entities is to remove these
> properties altogether rather than transfer them to an information item
> to which they do not logically belong.

We have already removed the [system identifier].  I think there would
be public outcry (I exaggerate slightly) against removing the
[character encoding scheme], which is I think regarded as a sort
of default encoding to use for output.

> I would also question [system identifier] and [public identifier] on the
> document type declaration info item.  The external subset is just
> another parsed entity.  If you are not providing system and public
> identifiers for all parsed entities, why provide then for the external
> subset?

I raised this some time ago (at that time we had an external subset
entity info item), and the prevailing view was that it was useful
for applications to be able to make decisions based on "the DTD".
Though it is technically "just another parsed entity" it is a
distinguished one from the user's point of view.

-- Richard

Received on Friday, 23 March 2001 08:01:35 UTC