draft-phillips-langtags-08, process, specifications, "stability", and extensions

[Harald Tveit Alvestrand]

--On 6. januar 2005 06:24 -0800 ned.freed at mrochek.com wrote:

>> I believe that John meant sect. 2.5 of RFC 3066, which does indeed
>> mention a matching algorithm.  However, the proposed changes in the
>> structure of tags interact badly with that algorithm.
>
> My reading of that text is that it goes out of its way to try and avoid
> direct discussion of a matching algorithm, talking instead about "rules"
> and "constructs". I no longer recall the circumstances behind this, but
> my guess would be that talking about algorithms directly moved this
> specification a bit too close to implementation work, which in turn would
> argue for the normal standards track and its ability to assess interop
> status, not BCP.

I lifted that stuff from RFC 2616 (HTTP/1.1), which is at Draft. The reason 
for putting it in 3066 was that people wanted to refer to it without having 
a normative dependency on HTTP.
The fact that it doesn't call itself an algorithm is probably mostly an 
effect of the writing styles of the various authors involved.

If we were splitting 3066 into a standards-track format specification and a 
BCP registration procedure, this should certainly go into the format 
specification - but I would argue that the RFC 3066 section 2.3 algorithm 
has sufficient interoperability experience and Internet-scale deployment 
that the correct status for it is Standard, not Proposed.

                   Harald