Lists Home |
Date Index |
On Mon, Nov 22, 2004 at 05:26:52PM -0800, Dare Obasanjo wrote:
> So detecting duplicate attributes and checking if characters are legal
> isn't well-formedness checking? I'm confused, how are you quoted out of
It's part of it of course, but not all of it.
To argue in your style, "so a subset is always equal to the whole?"
> 1. Derek points out at that some well-formedness checking is expensive
> 2. Liam responds that parsers can skip them and trust the generator of
> the XML
> 3. I point out that this is problematic on the Web where consumers of
> XML usually cannot trust producers of XML
I didn't use the word "trust". An alternative is to use a format
in which some or all of those errors can't occur.
For example, imagine a format in which attributes are always sent
sorted. Now the cost of checking for a duplicate is greately
reduced, although not zero. In practice I'd want all namespace
declarations up front, but that's a side-issue. Imagine further a
schema-based format in which the possible attributes are supplied in
sequence, with the positions mapping to names. Now no test for
duplicate attributes is required, and yet there's no breach of
This sort of approach can be difficult with arbitrary Web services
when there's no schema for a payload, but there are other cases
where it can work fine.
People are already using binary potocols to transfer XML efficiently.
Microsoft is doing it too. The question is not whether people should
do this, but whether there's a single way of doing it, or even a
single framework for saying which method you're using it, that
will be useful enough and across enough industries that it'd make
sense for W3C to specify.
I don't know the answer to that question yet.
Liam Quin, W3C XML Activity Lead, http://www.w3.org/People/Quin/