Update patches versus full update RPM

[Leonard den Ottolander]

Hello William,

> As a point of reference, two of the three Fedora machines I run don't have a
> CD-Rom, and none of them were installed from CD media.

Since you want upgrade the package you probably have a reference rpm 
around, being CD, local copy or somewhere on your local network.

> How does this help the problem of an overworked download server, then?
> People on dial-up will get timeouts just as easily as people on broadband.

> Notice how few rsync servers there are on the mirror list?  Part of the
> reason is that rsync can put a much larger load on the server.  Any kind of
> decompression and/or recompression will just make that worse.

Yes, I thought of that just now. Rsync is not an option since it wastes too 
much cpu on the server side, so we should go for "smart patches" (do any 
needed de-/recompression on the client side). But the issues involved in 
rsync (rsyncabilaty of a compressed file, or better whether the used 
algorithm recompresses the patched archive unambigouasly (sorry for lacking 
vocabulary here)) are relevant to this issue.

I am not sure how rpms are compressed, gzip, bzip2? To patch binary files 
inside signed rpms rpm needs to use a compression algorithm that meets the 
above condition.

Or rpms could be compressed with gzip using the --rsyncable option, so 
xdelta's on the whole rpm are more efficient.

Bye,
Leonard.

--
mount -t life -o ro /dev/dna /genetic/research