Danek Duvall wrote: > We're operating under the assumption that if two files have the same > cryptographic checksum, then their contents are exactly the same, so it > doesn't matter that they're stored at the same pathname. (There's an > inefficiency right now that the file will be overwritten, but we'll > eventually get rid of that.) > > Now, if there are real concerns about collisions, then short of alleviating > those concerns by moving to a new hash algorithm, we'll have to individually > compare each byte of a file with a given hash with the one already on the > system. That can get expensive, particularly since we already have to read > through the entire file to get its hash in the first place, and then again > since from one version of a package to another, most files will end up with > the same hash, meaning that we'll end up doing a lot of collision detection > on files that are actually identical. > > It's probably worth quantifying that cost at some point, but it's a low > priority at the moment. > > Danek > _______________________________________________ > pkg-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
We could do the checking at publication time so that clients aren't affected, since publication is so much less frequent. This will work unless multiple repositories (not mirrors) provide different versions of the same files... - Bart -- Bart Smaalders Solaris Kernel Performance [EMAIL PROTECTED] http://blogs.sun.com/barts _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
