Re: Requesting the RDF MIME type of an image

Hi all,

here we reach an interesting limit of the intuitiveness of the notion of
informational resource.

Le 23/10/2009 17:17, Toby Inkster a écrit :
> On Fri, 2009-10-23 at 15:02 +0100, Mischa Tuffield wrote:
>> Which you can perform content negotiation on, so that if you :  
>> request "Accept:application/rdf+xml" you would get back RDF  
>> and if you request html you would get back an HTML doc, perhaps an
>> html page with the image, and a human readable representation of all
>> of the metadata.  
>> and if you request image/jpeg (or whatever the correct MIME type is
>> for a .jpg file) you would get back the Image itself.   
>> This would allow you to change the file format of your picture if ever
>> need be (i.e. from .jpg to .png for example), keeping the URI of the
>> image constant and neutral to file format. 
> 
> I don't think this is a sensible way to use content negotiation. An RDF
> file and an image are probably not representations of the same resource;
> so they should not share a URI. (With the possible edge-case of an image
> which is the visualisation of an RDF graph.)
> 
> Perhaps:
> 
> 	GET /images/example HTTP/1.1
> 	Accept: image/png, image/jpeg, image/*;q=0.5
> 
> should return the JPEG, but:
> 
> 	GET /images/example HTTP/1.1
> 	Accept: application/rdf+xml
> 
> should return a 303 See Other to a different URL (e.g.
> </data/images/example> or </images/example;about>) which would provide
> an RDF description of the image.

Although I would tend to agree with Toby on this, I keep thinking both
options (303 or 200 for the RDF version) are valid.

After all, the pixel matrix conveyed by the JPEG encoding is also a
*representation* of a more abstract resource that we call an image (the
same *image* could be represented by a slightly different matrix using
another encoding, another color scheme, another resolution...).

So why would the JPEG representation be more *intrinsic* to the image
than the RDF representation. What if the RDF representation contained a
property
  :hasPixelGrid 'A65E8F9B87X78964...'
?

This is hair splitting, but I think it is worth questionning the
distinction between informational and non informational resources. This
distinction is often taken for granted, and I think it is often not as
clear as it seems...

  pa