Re: ETags vs Variants, was: Revising RFC2616 - what's happening

sön 2006-11-05 klockan 14:00 -0800 skrev Lisa Dusseault:

> There are certainly more creative alternatives.  For example, if the  
> spec said roughly that
>  - Vary headers are deprecated in the sense that servers implementing  
> this RFC SHOULD discontinue use of the header, and instead simply use  
> "no-cache"

Which isn't far from deprecating caching of HTTP in general, not only
the Vary header. The amount of Vary:ing content served today is quite
large and rapidly growing thanks to the deployment of content-encoding. 

So I would not see this as an option.

Vary is kind of usable today, problem with the specs is mainly
interoperability issues resulting in sub-optimal performance when both
sides doesn't agree entirely on how to read the specs. The result is
always correct as long as they conform to their own interpretation of
the specs..

The other problem is that many forget that they SHOULD send Vary at all,
but that's a different story.. changing it into another SHOULD isn't
likely to help that situation..

The biggest mess wrt Vary is the widespread confusion about the meaning
of ETag and entities / entity variants which is tightly coupled with
Vary.

Regards
Henrik