G'day, I've got an implementation of client-side delay pools for Squid-2.HEAD which allows for delay pools to be applied to the -client- side of a connection rather than just the server. It lets you shape the speed of replies to clients whether they're in the cache or not. I've also implemented a "class 5" delay pool which is specifically a per-client-connection rate limiter with an aggregate pool. I've also added in new config options to let you match on both the request and the reply when deciding which delay pool to assign to a given connection. I'll eventually work out how to implement server-side reply ACLs for delay pools which will let users match on MIME types when assigning the traditional delay pools; some of the work in this branch makes implementing that option easier. If you'd like to give this a whirl then you should check out the 'delay_pool_write' branch from http://devel.squid-cache.org/ . Instructions on how to use CVS are available there. Those who are interested in the gritty details may be interested in the changesets - http://devel.squid-cache.org/changesets/squid/delay_pool_write.html . I'll bring this work into Squid-2.HEAD once its had a bit more production testing. (This is where you all come in... :) It'll then be included in the upcoming Squid-2.8 release. I'd obviously appreciate any testing which the Squid community is able to help me with. All of this particular work is being funded by a generous customer with a specific need for the above. Thankyou to them for sponsoring this work! Adrian
