Re: restricting sex oriented material

From: Oskar Pearson <>
Date: Thu, 5 Jun 1997 10:09:13 +0200 (GMT)

Steve Green wrote:
> At 10:53 4/06/97 +0200, Oskar Pearson wrote:
> >Martin Gleeson writes:
> >Just on a further note: a group called basically does this as
> >a job, hiring people to go searching the net for this trash (interesting
> >job ;) so that they can offer it as a service to schools and so on.
> >The use a modified squid to handle the number of sites they block.
> >
> >They have a list of 180 000 sites that have "dirty pix/bomb-details" etc.
> 180,000 regex entries would grind your proxy to a screaming halt!
Nope - this used some kind of hash system
(this is from the archives - do a search for
> As far as I know all ACL's are dynamically created.

There is no hard-coded limit. However, I doubt very much that without a
lot of code changes you can put big lists into squid. In fact, we've
verified this with practical experience... :-).

Splay trees are just a fancy way of building binary trees. Since acls are
not dynamic once they have been loaded, splay trees are a waste of time.
In addition, simple binary trees are a very inefficient way of searching
URLs. They will certainly be better than a linear linked list, but not
fast enough for a large list on a production server. There are ways to do
very fast searches on large lists, but they require careful attention to
the problem space and result in extensive changes to the code.

> Also, wich acl do you have thet with help wit this.

I have completely different ACL types in my version of Squid (they are
optimized for the types of blocking patterns we generate). Again, these
changes to Squid are part of the service we sell, so I can't release
those, either.

Received on Thu Jun 05 1997 - 01:11:13 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:35:22 MST