Hi, Amos > Ah, well. *requiring* things beyond the RFCs (agreed compatible standards) > is a well known cause of failure. Just ask anyone trying to use NTLM on a > public website. Is "multiple POST requests in a http connection using keep-alive" RFC violation ? --mkishi On Wed, Sep 8, 2010 at 3:47 PM, Amos Jeffries <squid3@xxxxxxxxxxxxx> wrote: > On 08/09/10 17:46, Mikio Kishi wrote: >> >> Hi, Henrik >> >>> fre 2007-04-13 klockan 22:25 +0200 skrev Sebastian Weber: >>>> >>>> Hendrik, thank you - that solved the problem. Now, I have a new one >>>> however: >>>> >>>> The client -> proxy connection is kept alive correctly. The proxy -> >>>> server connection, however, is not. Squid closes the connection to the >>>> server after the client issues a second request. After that, Squid opens >>>> a new connection to the server and sends the second request over a new >>>> connection. >>> >>> For POSTs yes. This is a feature, not a bug. >>> >>> The reason Squid does this is because it can not be certain that the >>> server really supports persistent connections or that it is willing to >>> accept the request, and terminating a POST mid-way due to the server >>> closing the connection is very bad.. because of this Squid do not reuse >>> connections for POST requests instead opening a new connection. >> >> It's still a feature, right ? > > Mostly. We are working on HTTP/1.1 features to avoid closing them. A lot of > servers still require it though. > >> On the Internet, there are some web applications which requires >> multi post requests in a http connection... > > Ah, well. *requiring* things beyond the RFCs (agreed compatible standards) > is a well known cause of failure. Just ask anyone trying to use NTLM on a > public website. > > The IP layer itself requires that every application using it to communicate > be able to cope with connections being closed early. > > Amos > -- > Please be using > Current Stable Squid 2.7.STABLE9 or 3.1.8 > Beta testers wanted for 3.2.0.2 >
