Re: Bug 7034

On Tue, 23 Mar 2010, Sam Ruby wrote:

> On 03/23/2010 05:14 PM, Maciej Stachowiak wrote:
>> 
>> On Mar 23, 2010, at 7:24 AM, Sam Ruby wrote:
>> 
>>> 
>>> The following is a suggestion that I don't expect will not initially
>>> be popular, but I will put it out there in the spirit of
>>> brainstorming. It truly is a "lets turn lemons into lemonade"
>>> suggestion. I ask that everybody treat is as such.
>>> 
>>> There is a sincere desire by some people to require ampersands to be
>>> escaped, quote all attributes, close all open tags, get rid of tags
>>> such as acronym, and to rid the internet of the scourge that is
>>> presentational markup.
>>> 
>>> At the same time, the discussion about "this is XHTML" vs "not it is
>>> not" is showing no signs of going away. This discussion even persists
>>> when the alleged XHTML is served as text/html, does not conform to any
>>> known schema or DTD, and even when is not well-formed. I think that we
>>> have an opportunity to change the topic.
>>> 
>>> One possibility is to change the definition of the xmlns attribute on
>>> the html tag from being a talisman to an opt-in to best practices.
>>> 
>>> One downside of such an approach is that it would provide any means
>>> for people who author content intended to be served as
>>> application/xhtml+xml to opt out.
>> 
>> Another downside is that many people who want to "opt in to best
>> practices" would not agree that including the string
>> 'xmlns="http://www.w3.org/1999/xhtml/"' in a text/html document is
>> itself a best practice. If you want to propose multiple validator modes
>> triggered by something in the document itself, I would suggest using
>> something less potentially polarizing as the trigger. That would also
>> address the downside that you stated.
>
> I'll give my normal response to such assertions: to ask somebody to come 
> forward and state that such an approach is not acceptable to them personally 
> (i.e., I'm not looking for somebody to argue on behalf of unnamed others), 
> and to propose an alternative.

I believe this approach is a bad design (even by the standards of HTML). 
Making document conformance depend on the presence or absence of 
a superficially unrelated string seems confusing and error-prone. The 
choice of xmlns in particular seems bad as it conflates issues of XML-ness 
and conformance. This might make sense if the additional conformance 
criteria amounted to making the document a true polyglot document but, 
from the discussion so far, they do not.

The use of a magic talisman to change validator behaviour is, of course, 
similar to the HTML4 idea of strict/transitional doctypes. That clearly 
failed its original purpose (transitional doctypes are still widely used 
and so were not transitory). However compared to this proposal, the HTML4 
approach had the benefit of being more explicit; it is easier to grasp the 
concept of targeting a "strict" or "transitional" level of conformance 
than a "with xmlns" or "without xmlns" level of conformance.

I also note that, for authors who are uninterested in XHTML but are 
interested in strict conformance, this proposal negates one of the most 
popular features of HTML5; the minimal amount of impossible-to-remember 
boilerplate cruft needed to write a document.

As an alternative, I suggest the status quo: a single conformance level 
with the onus on validator authors to make their product usable for people 
that knowingly use obsolete features.

Received on Wednesday, 24 March 2010 08:23:23 UTC