On Tue, Nov 13, 2007, Alex Vorona wrote:
> >Another path would be to add another http_port flag making intercepted
> >requests on that http_port always use the original destination IP and
> >include that in the cache key. This smells more secure, but will not be
> >very good for the cache..
> Why not good for cache - because in most cases original destination IP
> will be in the set of IPs returned by DNS and will just slowdown cache
> with bigest key? There is no possibility to distinguish, how client
Yup. And you'd end up with quite possibly lots of different versions
of "stuff" and thus less caching.
> obtained IP-address - from hosts-file or from DNS. What if client
> forces in hosts only one IP(in the set of IPs returned by DNS) for
> multi-IP site - squid can't ensure that all requests will be done to
> that IP or cached from that IP. Then each multi-IP and cacheable site
> will have copies in cache by IPs count.
> Problem is not so simple, is I think before :)
I thought about that - do the DNS lookup, see if the IP the client is connecting
to is in the returned DNS set Squid would try to connect to, and cache the content
if it is.
It only works if the site returns multiple A records for the URI host; it
won't work for sites which do GSLB type tricks to reply with different IPs
based on "stuff" (load balancing, geographic/topological distribution, etc.)
Sure it can be done though. :0
Adrian
-- - Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support -Received on Tue Nov 13 2007 - 04:06:11 MST
This archive was generated by hypermail pre-2.1.9 : Sat Dec 01 2007 - 12:00:02 MST