Say I'm working with XHTML, which I intend to process somehow in
various ways (including XSLT, to make this somewhat on-topic) and I
include an appropriate DOCTYPE decl at the head of the source file.
A consequence of this is that default attribute values get added to
the tree before my processor has a chance operate on it. So now I've
either got to remove the default attribute values, which seems
ridiculous, or leave them in and send them explicitly to my clients,
which is pointless since they are, after all, defaults.
This happens even if I've explicitly disabled validation.
I can work around this by not using the XHTML DOCTYPE decl. This
means I've got to supply at least my own internal subset containing
any entities I might want to use, which is a hassle. Seems wrong,
since the source is in fact XHTML.
So, what am I missing here?