RE: End-of-line processing and attribute normalisation

> De: Richard Tobin [mailto:richard@cogsci.ed.ac.uk]
> Date: vendredi 11 février 2000 07:45
>
> Ok, I understand your points.  But now I wonder what the point of
> adding the phrase "after the end-of-line-normalisation..." was in
> erratum E24.

The point is to make sure the AttValue itself is line-end-normalized before
the normalization described in the bullet list is performed.  Say you have
in the document entity:

<elem att="foo
bar&baz;">

where the line end between foo and bar is #xD#xA and where &baz; also
contains a #xD#xA.  The paragraph before the bullet list ensures
normalization of the #xD#xA in the AttValue literal.  The third bullet takes
care of the #xD#xA in the baz entity (whether internal or external, 2.11
offers no guarantees in either case).

> Would it perhaps be better to remove that phrase, and instead add a
> note to bullet point three saying that it is the same process of
> line-end normalisation that is referred to in 2.11, and can therefore
> be achieved by the same pre-processing suggested in 2.11.

It would be better if 2.11 said "you must do that" instead of "this can be
achieved this way, or otherwise".  I do not know why 2.11 is like it is, but
there may have been good reasons.

> That is, a processor which follows the suggestion in 2.11 and
> normalises line breaks on input before parsing does not have to (and
> must not) do anything further to fulfill bullet point three.

Indeed.  Applying line-end-normalization early on all entities does the
trick, I think.  Parser implementers are allowed to do just that.

Do you still see something that could be an erratum?

Regards,

--
François Yergeau

Received on Friday, 11 February 2000 11:05:17 UTC