Re: store module abstractions

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Sun, 18 Feb 2001 21:43:36 +0100

Kevin Littlejohn wrote:

> Ok, that makes sense. Does that mean an incoming request for a file on a box
> with 5 different fs'es mounted would potentially have to query every fs to see
> if it has the file?

Depends on which kind of cache_dir distribution is being used. My
favorite for the time beeing is a modified CARP distribution.

> What happens in the case of collisions - say someone's mounted an fs
> previously used in another cache?

See above ;-)

Seriously, once one has made up his mind on how situations like this
should be resolved, making it behave in the desired way will not be very
hard. cache_dir selection is a very isolated piece of the code, so it is
easy to experiment with different algorithms there.

Note: There are two goals of all this restructuring being discussed:

* To make Squid a viable platform for research again, to allow trying
different approaches at many different levels.

* Bring the code up to a reasonable performance level, given what is
known today on how to better design a networking server/proxy.

/Henrik
Received on Sun Feb 18 2001 - 13:46:34 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:32 MST