W3C home > Mailing lists > Public > www-international@w3.org > April to June 2008

Re: Language tag education and negotiation

From: Asmus Freytag <asmusf@ix.netcom.com>
Date: Wed, 30 Apr 2008 10:45:24 -0700
Message-ID: <4818B034.9040903@ix.netcom.com>
To: Martin Duerst <duerst@it.aoyama.ac.jp>
CC: Andrew Cunningham <andrewc@vicnet.net.au>, www-international@w3.org

On 4/29/2008 10:25 PM, Martin Duerst wrote:
> At 12:40 08/04/27, Asmus Freytag wrote:
>> On 4/25/2008 11:45 PM, Andrew Cunningham wrote:
>>> basically no one size fits all, there needs ot be flexibility.
>> As long as we are discussing deficiencies of the language preference mechanism as currently implemented, here's another one:
>> For people who are bi- or tri- lingual (in the case of mutually un-intelligible languages) there's another situation that can crop up, which is entirely not handled by the current scheme.
> Actually, it IS handled by the current scheme, with what's called
> q-values. Q-values are used to indicate relative preferences from
> the client side, and relative quality on the server side. So a
> server could give original documents a q value of 1.0, and translated
> documents a q-value of e.g. 0.5 (q-values are always between 1.0 and 0.0).
I knew about q-values on the user side, but it would be their use on the 
server side that would allow support for the kind of scenario I had in 
mind. This leads to further questions. Are there any servers that 
support the q values? How does one set things up so a server can issue 
the correct q value?  And finally, is there an agreed upon convention on 
how to use such values?

It would take positive answers on all three of these to make it possible 
to actually utilize them.

Do the q values of sender and recipient get multiplied before 
comparison, or how does this work?
> q-values are sent by some browsers, which translate a preference
> list of e.g. en, ja, de, fr into something like
>    en, ja;q=0.9, de;q=0.8, fr;q=0.7
> or so. Apache also understands them. But they are difficult to set
> up, so they are not really used much.
Especially as browsers don't seem to allow several languages of equal 
status, nor large gaps. For example, assume the 'real' q values in your 
example were

	en, ja;q=0.9, de;q=0.4, fr;q=0.3

I would expect that such a user  might prefer a translation to 'en' of  
some 'fr' or 'de' document over the original, but would continue to 
prefer a 'de' over a 'fr' version of any document.

With correctly assigned q-values, and your convention of 0.5 for 
translation (and assuming multiplied values) this would work as 
intended. An 'en' or even 'ja' translation would trump an original in 
the languages with lesser preference. However, in the original scheme 
(fixed .1 offsets), you'd always get the original., even if they are in 
languages that, for the user in our example, might constitute languages 
of last resort.

Seems like this subject could use some of the educational effort that 
Leif was asking for.

> Regards,   Martin.
> #-#-#  Martin J. Du"rst, Assoc. Professor, Aoyama Gakuin University
> #-#-#  http://www.sw.it.aoyama.ac.jp       mailto:duerst@it.aoyama.ac.jp     
Received on Wednesday, 30 April 2008 17:46:13 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 21 September 2016 22:37:29 UTC