Re: URL-based ACLs

On Fri, Jul 17, 1998 at 04:23:28PM -0700, Andrew Daviel wrote:

> @Home hav a set of N head-end caches, not running ICP,
> and split webspace between them using an algorithm based on
> URL length (see http://proxy.rchmd1.bc.wave.home.com:8080 .. oh, foo
> .. see http://cr1001800-a.rchmd1.bc.wave.home.com/cache/home.proxy.pac,
> maybe. Anyhow, it contains some JavaScript such as
> function FindProxyForURL(url, host)
> {
> var num = (1 + (url.length % num_of_proxies));
> str = ("PROXY proxy" + num + ":8080; ");
> return(str);
> }
>
> If I select one of the proxies as a default parent (N=2 for me)
> I'll get a reduced hit rate on the head-end cache
> compared with someone using the .pac file in Netscape. I'm
> not sure whether they talk to each other, but they have (I believe)
> a common parent further up the net.
>
> Is there any way to hack Squid to include this kind of URL-based
> ACL (as opposed to host-based ACL) ?

While it's not as powerful as what you could do in the browser with
JavaScript, Squid 1.2 has cache_peer_acl, which you could use with
a url_regex ACL to do some simple switching among peers. I haven't
tried this myself, however, so somebody please correct me if I'm
mistaken.

-- 
Michael Fuhr
http://www.fuhr.net/~mfuhr/