Robert Collins wrote:
> To confirm my understanding:
> there should be only one entity instance for any given Varies
> combination?
Correct.
> (I.E. if varys is set to authentication, then potentially one instance
> per `authentication header`)
> of or varys is set to the content language header (sorry name escapes me
> just now), then potentially one entity per language header.
Actually one instance (headers and data) per ETag, and a map from
request header combinations selected by the Vary reply header to ETag
values. This because multiple request variants may end up in the same
entity variant.
If the reply does not have a ETag, then the map is from request headers
directly to entity, as there is no way to tell that two request header
variants maps to the same entity variant.
So it more or less adds two indirection layers on top of what we have
today.
/Henrik
Received on Thu Feb 22 2001 - 00:12:25 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:34 MST