Re: Handling aliased download sites

carson@tla.org wrote:

> If all the refresh data, etc. is in the URL, you avoid the issues. Of
> course, you now have to deal with decisions about when to free the URl entry
> (it's expired, but the content is still fresh from another URL, what do you
> do?).

Well, if you rewrite
        http://msvaus.../foo.cab
to
        \001multisite1\001/foo.cab
before taking the MD5, Squid won't even know that the object on the web
is a different object than http://mskyus.../foo.cab, because it hashes
to the same MD5. The normal rules governing last use and release apply.

Even if a given URI does not exist on a given server it won't matter
(give or take the negative cache time), as long as the content that is
there is guaranteed to be identical. For this purpose I'm thinking of
using
        http://ms....\.www\.connxion\.com\/cabpool/
as the base url. Stuff under Cabpool has generated names, apparantly
based on the MD5 of the content, so that if an object exists at all, it
is guaranteed to be identical to other copies.

Around the time of a release, IE is a big chunk of our bandwidth.

Cheers,

                                -- Bert

-- 
Bert Driehuis, MIS -- bert_driehuis@nl.compuware.com -- +31-20-3116119
Every nonzero finite dimensional inner product space has an
orthonormal basis. It makes sense, when you don't think about it.