Re: ISSUE-41/ACTION-97 decentralized-extensibility

On Tue, Oct 20, 2009 at 10:30 AM, Tony Ross <tross@microsoft.com> wrote:
> On Tuesday, October 20, 2009 at 12:28 AM, Jonas Sicking wrote:
>> On Mon, Oct 19, 2009 at 11:14 PM, Tony Ross <tross@microsoft.com> wrote:
>> > On Saturday, October 17, 2009 at 11:34 PM, Jonas Sicking wrote:
>> >> On Sat, Oct 17, 2009 at 4:54 PM, Tony Ross <tross@microsoft.com> wrote:
>> >> > I agree that XML Namespaces is more complex than simply using tag names
>> >> > for identity, but I'm curious as to what specifically makes you think it is too
>> >> > complex for the web.
>> >>
>> >> I've mentioned them before in this thread. But in short:
>> >>
>> >> The identifying name is a tuple rather than a string (arguably this could be
>> >> fixed by changing DOM/CSS, but so far such proposals have not been made).
>> >
>> > I see this as an issue broader than HTML 5. Perhaps such a proposal
>> > would be a good idea for the next versions of the DOM/CSS standards.
>> > Should HTML markup support namespaces, it would certainly be able to
>> > benefit retroactively from such new APIs.
>>
>> Well, I agree that namespaces being messy in the DOM is a bigger
>> problem than HTML. However HTML has so far skirted the issue by not
>> using namespaces.
>>
>> I'm reluctant to add a feature that hinges on that we'll improve API
>> elsewhere later, as that might never happen. I'm extra reluctant in
>> this case since I don't really see how to solve it.
>
> HTML has not skirted this as namespaces are already integral to HTML 5. The sole limitation is that you can't explicitly declare namespaces in markup.
>
> Examples:
> 1. Implicit namespaces for HTML, SVG, and MathML are generated by the current HTML 5 parsing algorithm.
> 2. HTML 5 allows dynamic creation of namespaced elements and attributes via the *NS methods.
>
> I want to make it clear that no new APIs, complexity, or problems are being introduced from a DOM perspective.

This is the type of slippery slope I'm really trying to avoid. I think
XML Namespaces are too complex, especially for the set of problems
it's trying to solve. We were forced to use them a bit since we wanted
vector graphics and math support in HTML. It was IMHO very unfortunate
that we were forced to either choose to create our own new spec, or
reuse SVG/MathML and which comes with namespaces. But that was the
situation and we chose the lesser evil of reusing those specs.

Adding the ability to declare and use namespaces doesn't just add
that. It'll also mean that more people will use it, and potentially
that more specs will use them. That is the fact that concerns me the
most (beyond the possible web compatibility issues).

/ Jonas

Received on Tuesday, 20 October 2009 21:35:37 UTC