[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: "Binary XML" proposals
- From: "Julian F. Reschke" <firstname.lastname@example.org>
- To: Al Snell <email@example.com>, Joe English <firstname.lastname@example.org>
- Date: Wed, 11 Apr 2001 09:56:18 +0200
> From: Al Snell [mailto:email@example.com]
> Sent: Tuesday, April 10, 2001 8:42 PM
> To: Joe English
> Cc: firstname.lastname@example.org
> Subject: Re: "Binary XML" proposals
> The way I'm thinking of it is as representing an element or attribute name
> as a pair of pointers, one to the namespace URI and one to the element
> name... that means two pointer comparisons for identity analysis (and note
> that the string used is the URI rather than the prefix; using the prefix
> instead of the full URI is just XML's way of doing the string-compressing
> trick anyway, and we don't need a second layer of compression :-)
Why don't you combine both in one scalar value -- in almost all cases, you
want to compare tupel (namespacename, name) anyway.
> Correct me if I'm wrong - there is absolutely no semantic significance in
> the choice of namespace prefix, right?
> <foo:hello xmlns:foo="asdf" />
> ...identical to...
> <bar:hello xmlns:bar="asdf" />
> ...my encoding will discard the prefixes foo and bar; will this in any way
> ever matter, even slightly?
Wrong. Namespace prefixes might occur in attribute values and need to match
the prefixes in the context they appear in (take XSLT and XSD as example).