Re: Question on the boundaries of content negotiation in the context of the Web of Data

On Feb 18, 2009, at 9:48 AM, ashok malhotra wrote:

> Jonathan, you said
>
> "I would think that CN is used (and intended to be used) not just for
> choosing between semantically equivalent entities, but also for
> semantic subsetting, such as abbreviated representations for mobile
> devices, low-resolution displays, audio vs. written, etc. Subsetting
> is certainly *not* equivalence."
>
> So, not equivalence but derived from?  I'm wondering how far we can  
> push this.
> Can CN we used to select say between a picture of a house and a text  
> description?
> I was told NO, but perhaps we are rethinking this.
>
> All the best, Ashok

Depends on who you ask.

- HTTP places no easily discerned restrictions on what you are  
permitted to do with CN, so people have done all sorts of things. If  
there are any restrictions they are of the SHOULD kind and not of the  
MUST kind, like everything else that it says about REST.

- Xiaoshu (who is probably not alone) wants everyone to stay away from  
giving advice to restrict CN in any way, under the assumption that the  
CN mechanism is too useful to have it suffer restriction and it's none  
of our business anyhow (sorry Xiaoshu, I'm paraphrasing). AWWW 3.5.1  
might seem to support this (sufficient consistency is up to the  
provider - i.e. do what you like), although it could also be read as a  
request to do something sensible (fess up to what sufficient  
consistency means to you, and then make sure you adhere to it).

- Tim has said use of CN to choose between representations that are  
"too different" is harmful (to some objective) and in my earlier  
message in this thread I tried to channel him. The TAG has concurred  
with Tim on this, e.g. in its advice to the authors of the Cool URIs  
for the Semantic Web note, even in the absence of any clear rule.

- I don't really have a strong opinion (I generally advise people to  
stay clear of CN whenever possible). I'd just like to see either  
clarity or a decision to abandon the project of formulating advice on  
this. I talked about one 'representation' being derived from (or  
semantically a subset of) another in the message you quote, but this  
is not a meaningful restriction in any strong sense, since given two  
representations A and B you can always posit a third representation C  
having the property that A and B both derive from C. So Tim has to be  
talking about something much more more nuanced than either  
"equivalent" or "derives from", and the objectives of any discipline  
(voluntary, obviously) will have to be teased out.

- My answer to Michael's example is that the server is clearly doing  
something wrong with entity tags, since a request to a cache for a PNG  
shouldn't be satisfied by RDF; but that its use of CN is within spec  
(HTTP), celebrated by Xiaoshu, and lamented by Tim.

So take your pick.

Today's message from Phil Archer reminds me that this issue may bear  
on the work of the Mobile Web Initiative, so it is probably timely.   
[1,2]

Jonathan

[1] http://www.w3.org/mid/499C2BAF.1010005@philarcher.org
[2] http://www.w3.org/TR/mobile-bp/#tc

Received on Wednesday, 18 February 2009 19:14:38 UTC