repository creation

[Jeff Johnson]

seth vidal wrote:

>After all this absolutely enjoyable discussion of repodata format I
>figured I'd mention something that would help users tremendously to
>reducing the amount of data they might have to download.
>  
>

Releasing packages to repo's every other day, or even weekly, would 
reduce the amount
of data as well, as rpm-metadata size is only a small part of the 
problem, too many
package rebuilds with teensy changes is the real problem imho.

But yes, splitting packages by usage category, like -debuginfo, and 
srpm, would
only help. Another split might be attempted by "popularity" as objectively
measured by number of downloads. "Popular" or "important" packages might be
released more often, while other packages less often, another way to reduce
the amount of information that flows from repos.

Another optimization would be to maintain rpmdb-fedora incrementally
on the client, as that's yet a 3rd time that headers are downloaded,
and maintaining the yum headers and rpmdb-fedora incremenatlly n parallel on
the client. That isn't a hard implementation to do.

Not carefully "number of downloads" lest fedora-devel traffic start to rival
the bandwidth used to download rpm-metadata discussing the relative
importance of packages again again.

<sensible ideas snipped to save band width>

73 de Jeff