Re: #428 Accept-Language ordering for identical qvalues

["Eric J. Bowman" <[e-mail]>]

James M Snell wrote:
>
> Well, considering that the http/2 discussion has already touched on
> the introduction of stateful compression, a potential switch to
> binary-header values, elimination of various elements such as response
> status-text and the host header, and so on, a discussion of
> eliminating conneg wouldn't be too extreme :-) ...
>

Nor would updating the WG charter to account for such changes ;-) ...

>
> The one thing to consider is that it ought to be at least possible to
> deprecate conneg without removing it entirely. We'll need to keep the
> mechanism around for http/1 interop and passthrough but we can say
> instruct developers that conneg ought to be avoided and we can
> discuss and highlight the appropriate alternatives.
> 

Which is exactly why I'd like to see this discussed further on the REST
list.  If the "solution" to conneg is to either require an extra round-
trip per request to indicate "compressed representation, please," or to
make compression stateful, then it's harder for me to buy into the
notion of server-driven negotiation as a "revolting feature."

IOW, I can't participate in a discussion about the way forward, if I
don't understand the problem with the status quo.  What am I missing?
I do think such theoretical architectural discussion belongs elsewhere;
at least as I see it, this list should be nuts-and-bolts protocol
writing.  I'm told this has been discussed before, links would help.

-Eric