RFC defining "Content Negotiation" (was: RE: Regrets)

Rob,

When you quote „the IETF RFC“ it seems that you quote the definition from RFC 2616 §1.3 [1]. That RFC is obsoleted through RFC 7230-7234. RFC 7231 §3.4 [2] offers a much deeper introduction to the topic including descriptions of “proactive negotiation” (a.k.a server-driven) and “reactive negotiation” (a.k.a. client-driven).

I don’t know (yet) how to address that in the prof-conneg document…

[1] https://tools.ietf.org/html/rfc2616#section-1.3

[2] https://tools.ietf.org/html/rfc7231#section-3.4


Best,

Lars

From: Rob Atkinson [mailto:rob@metalinkage.com.au]
Sent: Wednesday, November 14, 2018 6:53 AM
To: Simon Cox (E&M, Kensington) <Simon.Cox@csiro.au>
Cc: Annette Greiner <amgreiner@lbl.gov>; pedro winstley <pedro.win.stan@googlemail.com>; Dataset Exchange Working Group <public-dxwg-wg@w3.org>
Subject: Re: Regrets

Asking for a list of options and picking one is still arguably "negotiation" - although "mediation" might also apply. the IETF RFC merely states

" content negotiation
     The mechanism for selecting the appropriate representation when servicing a request."

It does not explicitly require support for lists of choices.

This description https://en.wikipedia.org/wiki/Content_negotiation#Agent-driven explicitly includes the case where a list is provided and the client chooses - but I don't see where a specification that defines this is referenced.

Then there is the sense that QSA may be multi-valued - there are no examples given, but this fits the sense of a list

And finally there are the issues which do need an issue raised IMHO,  about use of wildcards and returning a more specific profile given identifier of a general one. (these both occur in MIME type conneg, you dont have to specify sub-profiles of MIME types, but you may receive one.)







On Wed, 14 Nov 2018 at 15:53, <Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au>> wrote:
Thanks for clarification Annette. Indeed, I had not picked up the distinction.

From: Annette Greiner [mailto:amgreiner@lbl.gov<mailto:amgreiner@lbl.gov>]
Sent: Wednesday, 14 November, 2018 15:24
To: Cox, Simon (L&W, Clayton) <Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au>>; pedro.win.stan@googlemail.com<mailto:pedro.win.stan@googlemail.com>; public-dxwg-wg@w3.org<mailto:public-dxwg-wg@w3.org>
Subject: Re: Regrets


I'm distinguishing between negotiating for content and requesting specific content. I've not seen negotiation handled this way (where one indicates a list of acceptable options and gets a response chosen by the server), though I admit I may just not be aware of implementations. File-type suffixes allow selection of a particular file; they are not negotiation. Typical REST implementations provide another way to select specific files with a URL string as well, but those are not negotiations either.

On 11/13/18 8:49 PM, Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au> wrote:

>  Enabling multiple ways to handle content negotiation doesn't seem like a win to me, as using query strings is not a standard way of doing content negotiation otherwise, so that is a departure from current conneg standards.

Conneg using paths other than HTTP headers has been a common practice. File-type suffixes are the most obvious pattern. The Linked Data API that was developed primarily in UK government circles introduced some ‘standard’ QSA keys like _format, _view, _metadata quite a few years ago. It is true that they didn’t get adopted as a standard and I guess you could argue that this was because the idea was flawed, but it was responding to a clear need which I don’t think has gone away.

From: Annette Greiner [mailto:amgreiner@lbl.gov]
Sent: Wednesday, 14 November, 2018 13:36
To: pedro winstley <pedro.win.stan@googlemail.com><mailto:pedro.win.stan@googlemail.com>; public-dxwg-wg@w3.org<mailto:public-dxwg-wg@w3.org>
Subject: Re: Regrets


I guess I never did it explicitly, but I meant to vote +1 for publishing the prof ontology.

Sorry, but for conneg I have to vote -1 until a couple of issues have been addressed.

We have two outstanding substantial issues with the conneg doc that I would like to see at least marked prominently. One is the presentation of the QSA stuff as normative. It should not be normative. Alejandra pointed that out quite a while ago, I believe, and I agree. The conneg doc is about a standard for header-based content negotiation, and I think it is beyond our charter to give normative requirements for a QSA-based approach to conneg. I thought this was agreed by the editors, but the document still treats that section as normative.

In addition, I opened a second issue about the use of QSA to specify a second way of conducting content negotiation rather than as an example of how to enable discovery and selection of profiles by using query strings (#544).  I strongly supported a requirement for the latter, because it is necessary to enable human users to understand what is available and recognize when the data available are limited to a specific profile. Enabling multiple ways to handle content negotiation doesn't seem like a win to me, as using query strings is not a standard way of doing content negotiation otherwise, so that is a departure from current conneg standards. It may even be harmful, as it creates ambiguity as to whether content negotiation is available, since one would have to check both methods to determine that it was not available. Rather than addressing the issue of how negotiation obscures the choice of profile made behind the scenes for human users, it re-creates that problem in a new form. Finally, this additional approach is introducing new problems because it requires determination of how to handle situations where both types of negotiation are attempted.

-Annette

On 11/13/18 3:01 PM, pedro winstley wrote:
Hi Annette
Did you vote on the proposals for publication?

On Tue, 13 Nov 2018, 21:00 Annette Greiner <amgreiner@lbl.gov<mailto:amgreiner@lbl.gov> wrote:
Sorry, I won't be able to make today's meeting.
-Annette

Sent from my iPhone



--

Annette Greiner

NERSC Data and Analytics Services

Lawrence Berkeley National Laboratory




--

Annette Greiner

NERSC Data and Analytics Services

Lawrence Berkeley National Laboratory