W3C home > Mailing lists > Public > public-html@w3.org > March 2010

Re: Possible *third* proposal for ISSUE-41 Distributed Extensibility

From: Sam Ruby <rubys@intertwingly.net>
Date: Mon, 22 Mar 2010 09:20:13 -0400
Message-ID: <4BA76E8D.1030109@intertwingly.net>
To: Henri Sivonen <hsivonen@iki.fi>
CC: HTMLwg WG <public-html@w3.org>
On 03/22/2010 07:35 AM, Henri Sivonen wrote:
> On Mar 19, 2010, at 14:34, Sam Ruby wrote:
>> On 03/18/2010 11:12 PM, Tab Atkins Jr. wrote:
>>> On Thu, Mar 18, 2010 at 3:17 PM, Ennals,
>>> Robert<robert.ennals@intel.com>   wrote:
>>>> Comments?
>>>> If it seems that people might like this then Iíll write it up
>>>> formally.
>>> Not a fan.  *Anything* that uses an XML Namespaces-like mechanism
>>> for embedding new elements is a bad idea, imo, because the
>>> fallback story (necessary to activate an experimental feature in
>>> multiple browsers, and to transition from experimental to
>>> standardized versions) is so horrible.  Maciej argued that it may
>>> be worse than not doing it at all, and just using the public name
>>> from the start.
>>> This is simply unusable as a way to allow browsers to add
>>> experimental features without clashing with each other and future
>>> standardized versions of the feature.
>> [co-chair hat off]
>> I don't believe that this proposal should be used by browsers to
>> add experimental features.
> Do you believe that D.E. should be used by parties other than
> browsers themselves to add experimental features to browsers (e.g.
> via Firefox extensions that observe what happens in Web content)?

Two answers to your question: no, I don't think that's the primary use 
case, and as to that particular use case: it should be discouraged, 
guided, but not outright outlawed.

Longer answer on the second part: "parties other than browsers" is not a 
well defined set, and the guidelines for such parties should not be 
materially different than the guidelines for browsers.

The use case shouldn't be outright outlawed: SVG and MathML kinda mostly 
worked that way.  Here's another example that /might/ be able to work:


These cases exist, but are rare.  And generally take a lot of care. 
There are a few common patterns for such usage that should be explored: 
a container element, default namespaces, avoiding (mostly) existing 
element names.

But as I said, that's rare, and not the primary use case (though I 
understand the importance of probing the edge cases in discussions such 
as these).  As to the first question, longer answer here:


- Sam Ruby
Received on Monday, 22 March 2010 13:21:00 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 29 October 2015 10:15:59 UTC