Re: connection reset by peer messages

--MimeMultipartBoundary
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable

Chris Wedgwood wrote:

> I don't think this is the case for the IIS servers though. The server
> accepts an initial connection and then close it before sending _any_ da=
ta
> whatsoever. I'm not sure if this is before or after a request is receiv=
ed
> (that distcintion is kind a pointless anyhow when the request is smalle=
r
> than the window size).

This is exacly the case that part of the HTTP spec speaks about: How
to deal with a HTTP connection where the server closes the connection
before sending a response. If the close is detected as a "Connection
reset" or "No responce" does not matter, it is technically the same
thing and the difference is only a sideeffect of how TCP/IP works.

> Now, this may or may not be acceptable, but the fact is this happens an=
d
> that the squid emits a 'connection reset by peer' response when Microso=
ft
> Proxy II seems to retry these. (I've not actually confirmed it does oth=
er
> than I don't see errors that would indicite this when using it).

Squid should NOT emit these messages, it is a bug/misfeature/HTTP
violation.

> I agree that squids probably do the right thing and the client should d=
eal
> with it as appropriate, but then some people start whacking reload whic=
h can
> deminish the effectiveness of the cache.

Squid does NOT deal with it in a proper way if it emits a "Connection
reset"
error message on first try. Either it should retry the operation for a
undefined number of times, or leave it to the client to decide (by
closing
the connection without a reply).=20

---
Henrik Nordstr=F6m
Sparetime Squid Hacker
--MimeMultipartBoundary--