Re: Call for Adoption: draft-reschke-rfc54987bis

[Willy Tarreau <[e-mail]>]

On Tue, Mar 31, 2015 at 02:40:36PM +0000, Poul-Henning Kamp wrote:
> --------
> In message <[e-mail]>, Mark Nottingham wri
> tes:
> 
> >We discussed this document in Dallas:
> >  <http://tools.ietf.org/html/draft-reschke-rfc5987bis>
> >
> >Based on the feedback received, I believe that we should adopt this
> >document as a WG product, with a target of Proposed Standard.
> 
> Solving the problem:  Yes, good idea.
> 
> "Solving" it this way:  Bad idea.
> 
> First, we're worried about transmission times for HTTP so making
> the charset selection per header-subfield is a horribly inefficient
> way to solve the problem.

I think it still makes sense because some intermediaries could be each
adding one header field and it would really not be handy for them to
have to lookup a certain header to know what format to emit theirs.
However, maybe per-header could be enough. But I guess Julian wanted
to ensure that interoperability is the least possibly impacted, which
probably starts by not mangling the header value before the semi-colon
for cases which already work and whose encoding is "implicit".

> Second, do we really want to make it possible to have one subfield
> of a header be KOIR8 and the next subfield be codepage 1251 ?

I think that's a legitimate question.

> Third, are there *any* valid reasons to even allow other charsets
> than ISO-8859-1 or UTF-8 from 2015 forward ?

Idem. And if we don't need to do more than that, then probably we
just need a boolean to say "this is not ISO-8859-1, hence this is
UTF-8" and make the encoding implicit by the sole presence of the
encoding tag (eg: the "*" or "=", I don't remember right now).

Best regards,
Willy