Re: [comment] XMLHttpRequest Object - Address Extensibility

"Bjoern Hoehrmann" <derhoermi@gmx.net>
> * Jim Ley wrote:
>>This is very silly, the VendorMember scheme is entirely stupid, it's
>>completely useless for authors, we can't do anything with it, and is much
>>worse than simple invented terms that eventually get standardised.
>
> It is very useful if it is immediately obvious that some member is a
> possibly experimental and unsupported vendor-specific extension.

Why does this matter?  such a "immediately obvious" is only relevant for 
people who are simply cutting and pasting code, such people will care not 
about any standard features, they'll purely care about if it works - this 
means all UA's will get pressure to implement the VENDORFOO property, and 
the WG will get pressure to standardise the VENDORFOO property - you 
certainly don't want VENDORFOO and FOO to exist everywhere.  As soon as 
VENDORFOO is standardised (even only de-facto) this immediate obviousness 
disappears.

More "professional" developers, will be looking at the documentation, will 
be QA'ing their source, which will rapidly indicate the experimental nature 
of a feature.  UA's may want to indicate the experimental nature of some 
things in their "Error console" - it's full of everything else these days.

> There are many alternate ways to enable vendor-specific extensions
> to simply allowing vendors to make up new members as they like. As
> an example, a 'ext' prefix could be required which W3C promises to
> never use in future versions of the specification and vendors may
> to in that namespace whatever they like. This would mitigate the
> problem of multiple vendors adopting each others extensions.

It would not solve the transition between experimental and permanent.  It's 
also not clear why the WG gets to experiment in the whole namespace, but 
other groups do not - a PR with no implementations is every bit as 
experimental and likely to change as a UA's extension.

[The Registry idea...]

A registry doesn't suffer from the above problems, but of course suffers all 
sorts of other problems, someone's got to run it, someone's got to somehow 
prevent Jibbering Implementation grabbing all the useful names on a junk 
implementation, whilst still allowing a legitimate Jibbering Implementation 
to create names rather than just a W3 hegemony.

Cheers.

Jim.

Received on Friday, 21 April 2006 21:27:37 UTC