mån 2009-07-13 klockan 08:14 -0600 skrev Alex Rousskov:
> Have you shared your opinion on the practicality of that MUST with the
> 2616bis folks? It looks like the specs should be adjusted if others
> agree with the "client knows better" point of view in the proxy context.
> This would certainly simplify the long-term fix.
I think it was informally discussed some year ago, but not sure. Will
take it up again.
Chunked encoding in requests is still a bit of a special case when it
comes to the HTTP protocol, this due to the lack of negotiation before
the request is sent. It's quite understandable the specs is written the
way they are given this fact, but unfortunately a bit impractical to
implement in proxy environments for many reasons.
What can perhaps be done wihin HTTPbis is to relax the requirement
somewhat for proxies, and more clearly point out the interoperability
issues one can expect. But I feel a general pessimism in the
workinggroup when it comes to chunked requests so I am not entirely
convinced the discussion will fly in any direction..
When it comes to infrastructure proxies can be expected to know if their
peers support chunked encoding or not, but the same can not be said
about proxies knowledge of origin servers as both forwarding path and
time is fuzzy parameters. End-Clients needing to use chunked encoding in
requests can however be expected to know the servers they talk to quite
well, even if there is months between their requests.
Regards
Henrik
Received on Mon Jul 13 2009 - 15:08:03 MDT
This archive was generated by hypermail 2.2.0 : Tue Jul 14 2009 - 12:00:05 MDT