RE: Multiple request blocking

> > It is about the ability for new clients to join a already
> running server
> > request for the same object.
> >
> > Three options exists
> >
> > a) From start (wanted by Eric, as it takes time before the
> reply headers
> > are received as required by 'b')
> > b) Allowed once it is known that the object should be
> cachable (current
> > model)
> > c) Never

We could just have later requests wait until enough is known
(which maybe is what we're doing now). I agree with Andres
that joining without knowledge is a Bad Thing(tm).
Or, we could introduce something like CPU pipelining.
We optimistically join, but as soon as we can determine whether
the request can be shared (i.e. after getting HTTP headers from
the origin) we add an extra check and if it can't share we
un-join and start a new request.

-- 
	/kinkie