Re: POST-WITH-NO-SIDE-EFFECTS method

Duane Wessels writes [in reply to Jeffrey Mogul] :
> 
> >This suggestion (a new kind of POST method) has a certain elegance
> >to it.  What worries me most is that it is probably incompatible
> 
> I like the idea of adding a POST-W-N-S-E method so that we effectively
> make use of a caching hierarchy.  The Harvest cache doesn't forward
> POST requests to parent/neighbor caches (unless behind a firewall).
> Without the new method (or some other indication in the request)
> POST-type requests won't be able to take advantage of the hierarchy.
It's a good aprroach: why to bother the whole cache hierachy with non-idempotent
requests? They are the business of the client and the origin server.
Caches will only slow down the process if there is no chance to serve something
from their caches unless they do a firewall traversal too.
Of course this requires some knowledge of the network connectivity 
on the client side.
Oops, Duane already told this.
> I also hope that user agents will become more intelligent about
> what they do and do not send to a proxy/cache (and which one to
> use, etc.).  If a browser had that ability today, I bet it wouldn't
> forward any POST requests to a cache.
Some modifications in Forwarded: header can help to learn about
network connectivity.

Andrew. (Endre Balint Nagy) <bne@bne.ind.eunet.hu>

Received on Friday, 5 January 1996 09:14:29 UTC