Re: Content Sniffing impact on HTTPbis - #155

lör 2009-06-13 klockan 19:43 +0000 skrev Ian Hickson:

> In this particular case, I really would encourage the use of terms people 
> understand. In any case, the term "resource" is correct when refering to a 
> file, stream, or other "bag of bits". The terminology used by HTTP in this 
> instance is inconsistent with wider usage (and remarkably confusing).

HTTP resources vary over multiple axis, and may have a large number of
different representations each with different bags of bits but
supposedly mostly the same semantic meaning to the end user.

  - Time, as already mentioned (versions)
  - Language (English & Swedish, depending on preference by the user)
  - Encoding (i.e. identity or gzip depending on capabilities of the
agent)
  - Cookies (i.e. anonymous or authenticated variants of the same
resource)
  - and even media type (i.e. gzip / jpeg / png depending on
capabilities of the agent)
  - or other server defined properties

resource is used as the verb defining the resource in general,
irregardless of it's specific representation, and is identified by an
URI.

Each representation MAY also have a unique URI identifying just that
representation as a resource, but is kind of besides the point.

HTTP intends that these representations should have pretty much the same
semantic meaning to the user, but can naturally not enforce it. It's
clearly stated however that the representations DO NOT share the same
bag of bits..

We have already had this resource/entity/whatever discussion many times,
and settled for resource in the general term and representation (of a
resource) for a specific bag of bits. Lets not open that discussion
again.

Regards
Henrik

Received on Monday, 15 June 2009 13:07:42 UTC