Re: I18N feedback on Issue-88

fantasai, Wed, 19 May 2010 02:21:44 -0700:
> On 05/19/2010 02:06 AM, Henri Sivonen wrote:
>> "Addison Phillips"<addison@lab126.com>  wrote:
>> 
>>> - HTML should (continue to) strongly recommend the presence of @lang
>>> (and warn in validators if it is not present)
>> 
>> If validators did that, there'd be even more templates, etc., filling in a
>> placeholder value that doesn't necessarily have anything to do with the
>> actual content.
> 
> In that case, I would suggest following Leif's suggestion and only
> posting a warning about a missing lang="" if the Content-Language
> HTTP header or <meta http-equiv> pragma is present. This is more
> likely to catch authors who are trying to specify the language but
> doing so wrongly, and avoid the authors who don't care.

Don't you think that making the pragma an outright error *could* work, 
provided that the language fallback warning from my current change 
proposal is implemented as well?  I see 3 differences from Ian's "Make 
Content-Language pragma non-conforming" proposal:

	A. validators/spec would not ask authors to "use @lang instead";
	B. validators would show an additional warning in case the 
       pragma caused the fallback language measure to kick in;
	C. validators would warn if the C-L HTTP header caused language
       fallback; The warning for B. and C. should ask authors to 
       take control via html@lang.

Advantages:

  1. Limits the actual fallback- fallback in the wild is 
     more often caused by pragma than by HTTP.
     (MAMA showed pragma to be used 10 times as often as HTTP.)
  2. Naturally increases focus on @lang
  3. A step towards the ideal goal which the I18NWG expressed in
     their feedback: that we get rid of fallback.
  4. doesn't assume that authors misunderstand C-L.

Disadvantages:

   - It might be confusing that the C-L pragma may affect the document 
language despite being invalid - who will notice it? This is the why I 
suggest to show both a fallback warning (if it kicks in) as well as an 
error - to make authors aware. Such a warning can also limit the 
occurrence of bogus language tags.
   - Inability to validly implement "fallback" without a server.

Comments:

	A., B. and C. differ from Ian's proposal to make the pragma invalid.

	Regarding A.: to tell authors to "use @lang instead of pragma" assumes 
that all authors have the same misinterpretation about the semantics of 
the Content-Language pragma. It is better to remain completely silent 
on what to do "instead", rather than assuming any particular 
(mis)understanding. Eventually one should also list HTTP C-L as an 
alternative to the C-L pragma.
	Regarding B.: I am open to discussing to not implement (B), in case 
two error messages instead of only one would seem confusing.
	Regarding C.: An alternative to (C) could be to require UAs to not use 
HTTP as basis for fallback.
-- 
leif halvard silli

Received on Thursday, 20 May 2010 01:49:56 UTC