Connection presistance in Squid is implemented per RFC2616 where the
server connections are maintained separately from the client
connections. When a request has been received (from a persistent
connection or a new one, does not matter) Squid selects a connection
from the pool of idle connections to that server, or opens a new
connection if no idle connection is found. When the server has finished
sending the reply to Squid the server connection is returned to the pool
of idle connections for that server to allow reuse on the next received
request for that server, idependently of from which client the next
request is received.
Squid is slowly moving towards full support of HTTP/1.1, but this will
take a while yet (probably a year or so). However, connection
persistance is one of the few areas that is actually fully finished
already (if you discount the fact that Squid also supports HTTP/1.0
style persistent connections, which HTTP/1.1 specifications does not
allow).
Some time later we might also experiment with request pipelining even if
discouraged for proxies in the HTTP/1.1 specifications. However, doing
it correctly in a proxy is far from trivial, and is probably why RFC2616
discourages it's use in proxies.
-- Henrik Nordstrom Squid hacker phil_blake@uk.ibm.com wrote: > > I've looked through the FAQ database and can't find an answer to my > question. > > I have a client which sends HTTP/1.1 POST requests to a server, requesting > a persistent connection. Looking through the Squid trace, it creates a > socket > to the client and another to the server (lets call these socket A and B). > If the client > then makes a new request to the same server, Squid creates sockets C and D. > > Sockets A and B are still active and data is being POSTed over these > sockets. > Data is also being POSTed over sockets C and D. This works fine for a > while. > > At some point, data POSTed on socket A is sent over socket D to the server, > not > socket B, as expected. Is this WAD ? Is there a way to guarantee use of the > socket > pairs ? > > Finally, are there plans for Squid to fully support HTTP/1.1 ? > > Thanks for your help, > Phil Blake > MQSeries Interfaces Internal phone 246027 > Hursley, England External (+44) 1962 816027 > > Mail Point 211 Fax (+44) > 1962 849268 > > -- > To unsubscribe, see http://www.squid-cache.org/mailing-lists.html -- To unsubscribe, see http://www.squid-cache.org/mailing-lists.htmlReceived on Tue Sep 26 2000 - 15:57:00 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:55:29 MST