Re: OT: Rsync not effective at minor changed packages

[Brian May <bam@debian.org>]

>>>>> "Petr" == Petr Cech <cech@atrey.karlin.mff.cuni.cz> writes:

    >> This means rsync did practically download the full file, which

    Petr> it did. these are different files, so it download them
    Petr> whole. How should rsync know, that -0 and -2 are similar?

I can think of a solution:

<flameproofsuit>
Why encode the version in the filename at all?

While this will mean it is harder to manually identify what version a
deb package is, most people I think use apt now anyway. Also,
I have heard that limitations in apt have been/will be fixed, eg
not being able to keep track of stable and unstable packages
at the same time.

There are other ways the version number can be attached to the file,
too, for instance, if viewing the directory listing via Apache, you
could have a .htaccess file with "AddDescription" statements (of
course, you could also get more sophisticated then this, too).

Furthermore this will allow maximum bandwidth saving with protocols
like rsync and rproxy.

rproxy, and extension to HTTP, see:
http://linuxcare.com.au/projects/rproxy/ for a description (not sure
if this is the main site). It looks like they eventually want to
integrate it into Apache, Squid, and Mozilla, for instance. My
understanding is that it is better then rsync in that it allows
caching of files, where as rsync doesn't. Is anyone packaging the
stand-alone rproxy for Debian?

of course, I don't really understand the implications of using rsync
on compressed files and/or if it is practical for rsync/rproxy to
decompress them "on the fly".
</flameproofsuit>
-- 
Brian May <bam@debian.org>