Re: review of "The root element" subsection

[taking this onlist, since it's relevant to everyone and nothing  
confidential or private]
On Jul 11, 2007, at 10:31 AM, Andrew Sidwell wrote:

> Robert Burns wrote:
>> On Jul 10, 2007, at 6:27 PM, James Graham wrote:
>>> However it introduces two equivalent ways of doing the same thing,
>>> with only a minor benefit to the new approach. Dan has already  
>>> pointed
>>> out that this is contrary to several design principles.
>>
>> No, it introduces an alternate second way to do the same thing. My  
>> first
>> choice would be to not introduce a new way at all. My suggestion  
>> is that
>> given the fact that someone wants to introduce a new attribute, it
>> should be introduced to the <html> element and not the <meta>  
>> element.
>
> The attribute already exists -- that is, it's handled by UAs today,  
> just
> like the autocomplete="" attribute on forms.  So to dismiss it as
> "someone wanting to introduce a new attribute" is ignoring the  
> reality.
>  It is not the same as e.g. the video element, which had no UA  
> handling
> when proposed.

@charset is not yet an attribute in terms of author conformance. The  
draft proposes making it one. That's the "reality". I question  
whether that is a worthwhile thing to do: especially in light of the  
very narrow cow path we'd be paving (according to your next comment).

> [1] shows that the charset attribute was present in 0.5% of pages
> surveyed with a <meta> element (calculated by using the values for the
> size of rectangles in the SVG source).  The survey was of just over  
> one
> billion pages; 5% of that is five million pages.  One can conclude,
> then, that this is most definitely not a new attribute.
> [1] http://code.google.com/webstats/2005-12/metadata.html

How many of those present charset attributes were even relevant  
(i.e., how many were on pages that didn't have a BOM or an HTTP  
header that provided encoding information, etc)?

Charset is only currently an attribute in terms of UA conformance in  
the sense that some UAs (may) employ algorithms that sweep up  sloppy  
coding (found in only 0.5% of documents by the research). My  
suggestion on the wiki simply calls for some more robust research  
into whether that's actually true of any (or all) current UAs.

So I think its a stretch to claim that this is not an introduction of  
a new attribute. Its also a strecth to claim all of the necessary  
testing has been done. I've laid out some criteria for UA testing and  
I'm working on putting together some test that meet that criteria  
(obviously this can't be tested with the Live DOM viewer).

Take care,
Rob

Received on Wednesday, 11 July 2007 21:30:21 UTC