Re: URI: Name or Network Location?

On Jan 23, 2004, at 18:36, ext Benja Fallenstein wrote:

> Hash: SHA1
> Hi Sandro,
> I can't answer for Patrick, but I have my own internally consistent
> views of this, which I hope may be of interest :-)
> Sandro Hawke wrote:
> | While you're on the subject, and because I know you have seriously
> | consistent views on this: what does the word "representations" mean
> | there?  My gray understanding is the community on the whole has no
> | consensus here, so a not-well-defined term is used.  But I suspect it
> | is well-defined for you.
> For me, generally, a representation is something that the "owner" of a
> URI thinks is useful for a client to serve in response to a GET request
> on that URI. Usually, it is what the "owner" wants to appear in a
> browser when a user types the URI into the address bar (although this 
> is
> not the only use; an RSS feed will more often be consumed by an
> aggregator than shown in a browser).
> This definition is intentionally "backwards," i.e. based on the
> interests of the URI owner rather than on the direct semantic
> relationship between resource and representation. I believe that URI
> owners should be able to experiment with this.
> However, I think that almost all representations will fall in one of 
> two
> categories:
> - - Instance: The resource is a document of some kind, such as a 
> textual
> document or a photo. The representation is a byte sequence that encodes
> the resource in some digital format, which can be presented to a user. 
> A
> resource can have different instances, which can vary slightly (GIF vs.
> PNG -- same pixels, different format) or greatly (different 
> translations
> of the Bible). The document can also be "this particular JPEG file," in
> which there would be only one instance representation, which would be a
> bit-equal copy of the JPEG file.
> - - Description: The *representation* is a document that gives the user
> some information *about* the resource. The resource itself can be
> anything. For example, the RSS feed of a blog would be a possible
> description of the blog itself, as would be the blog's "About" page.

For what it's worth, this coincides with my own views fairly well.

> An interesting thing to note is that even digital documents can 
> usefully
> have 'description' rather than 'instance' representations. E.g. I would
> believe it to be useful if assigned
> as the canonical URI for identifying
> the Haskell report plus addenda &c. Putting the URI into a browser
> brings up a page that links to different HTML, PS, and PDF versions of
> the report.

Or better, IMMHO, allow agents to explicitly ask for a formal 
of any resource denoted by an HTTP resolvable URI per something such
as URIQA[1].

> | Try replacing "representations of" with
> |
> |    (1) "content (MIME Entities) associated with"
> I guess this can be read as being equivalent to my general definition,
> if you see "associated" as "whatever the URI's owner sees fit as an
> association."
> |    (2) "serializations of"
> This would be my 'instance' type, which I see as only one possible type
> of representation.

I think that it is essential to continue to allow the associations 
resources denoted by URIs and representations of those resources
accessible via those URIs as loose and free as possible, and yet
provide for an explicit, formal, and efficient mechanism by
which owners of URIs can unambiguously state the nature of
those resources, their representations, and the relationships
between them. I.e. something like URIQA[1].

This allows the web to continue to evolve in novel directions while
also allowing us to capture, where useful, the semantics of the
web at any given point in time.




> Cheers,
> - - Benja
> Version: GnuPG v1.2.4 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird -
> FlGO+abUK/nxwnsIoTvoAPY=


Patrick Stickler
Nokia, Finland

Received on Monday, 26 January 2004 03:02:24 UTC