Re: Conventions for Sharing User Agent Profiles

hallam@Etna.ai.mit.edu:
>
 [Koen Holtman:]
>>Transparent content negotiation does away with this limitation, and
>>feature negotiation exploits the absence of this limitation to the
>>fullest extent, yielding a framework which allows an open,
>>evolutionary approach to the problem of creating a shared language in
>>which to express capabilities and preferences.
>
>Agreed that these are all related I think we have to get the
>concept space right before we try to compress everything into
>one unified scheme.  

The goal of the current draft spec is not to define the ultimate
unified scheme.  I don't think it is possible right now to define a
unified scheme that will last without changes for, say, the next 10
years.  Once we get more experience with content negotiation, we will
discover social and technical problems we cannot even begin to imagine
now.

Quoting from the draft spec:

 19.3 Open issues in transparent content negotiation
 [...]
   - Though it is expected that the feature negotiation framework will
     solve many current and future negotiation problems, it is also
     expected that there will remain current and future negotiation
     problems not solved by feature negotiation.
 [...]

Feature negotiation is my best attempt at extrapolating the mechanisms
we will need from current negotiation experience.  If you see something
missing in feature negotiation right now, by all means say so that we
try to think up a fix.  We already identified the need to add
non-numeric feature tags, and they will be added in the next version
of the draft.

>One reason why I was skeptical of PEP was
>its "all encompassing" nature which never quite appeared to be
>grounded.

Transparent content negotiation makes no claim to be the ultimate
mechanism.  Transparent content negotiation tries very hard to be
extensible, because I fully expect it not to be the final answer.

>I think that there is a role for some king of multi-circuit 
>exchange protocol. I think that needs to be more general
>than simply content type or even feature negotiation.  I
>am currently looking at assertion exchange based on common
>reference terms which are essentially pure URNs except much
>of the baggage associate with URNs has to be lost so I'm
>using a new name.

Hm, assertion exchange sounds very much like what feature negotiation
does.  For feature negotiation, the common reference terms are feature
tags.

>I think we need to kick the requirements space arround a bit,
>get a feel for the constraints on implementation and then look
>at architecture again.

We have been kicking the requirements space on this list, and on the
content negotiation subgroup mailing list, for more than a year now.
Some of the concepts in the conneg draft have been kicked around since
at least 1993.  I don't think any more kicking can give us a much
better feel.

We cannot look much further from where we stand now, and we definitely
cannot see the final destination.  We have a draft spec that plots the
next few steps over the horizon.  Now is the time to finish the spec
and start walking.

>                Phill

Koen.

Received on Sunday, 11 August 1996 17:37:59 UTC