Re: i109: Clarify entity / representation / variant terminology

Mark Nottingham wrote:
> 1) "variant" occurs 16 times in the -02 specs, and a good portion of
> those is the "requested variant" text. If we ignore those instances
> for the moment, the expedient thing to do would seem to be to change
> the remaining occurrences to either "entity" or "representation," and
> remove this term altogether.

When deciding between "variant" and "representation," I think it makes sense 
to look at standards for extensions to HTTP to see how these two terms are 
being used. I looked at all the proposed standards that were obviously 
related to HTTP and WebDAV. "Variant" and "selected variant" is only used 
with the RFC2616-defined meaning. "Representation" is sometimes used to 
refer to variants, and sometimes it is used in its more general sense to 
describe other things. "Represent" and "represented" are often used in their 
more general sense as well. Details below.

Obviously, some of these proposed standards are not so relevant. However, to 
me it seems like a good idea to give "variant" a very precise meaning and 
let HTTP-related standards use "representation," "represents," 
"representing," and "represented" in their more general sense to describe 
other things. "Variant" also has an obvious relationship to the "Vary" 
header.

Defining "representation" in a way that isn't specific to content 
negotiation (without using "represent") may prove tricky as well. It seems 
to me it would be easier to remove the definition of "representation," leave 
the current definition of "variant" as is, and change all instances of 
"representation" to "variant" or "entity" as appropriate.

Regards,
Brian

RFC 3229 (Delta encoding in HTTP) uses "variant" consistently, and uses 
"representation" only in its more general sense (and to define "variant."). 
It uses "variant" consistently throughout.

RFC 4387: Uses representation once, but again it is in its more general 
sense.

RFC 3230 (Instance Digests in HTTP) uses representation and variant almost 
evenly, and tends to use "variant" when it is trying to be precise.

RFC 4918 (WebDAV): Uses "representation" once when "entity" would be more 
appropriate, and uses the interesting phrase "representation returned upon 
GET" a couple times. It never uses "variant."

RFC 2227 uses "variant" throughout, and uses "represents" in a more general 
sense.

RFC 2817 doesn't use either term.
RFC 2935 and RFC 2965 do not use either term, but it does use "represents" 
in its more general sense.

Received on Friday, 9 May 2008 16:41:50 UTC