[packagekit] Caching of repository metadata (esp. Yum)

[Jud Craft]

Hello there.  I wanted to make a criticism and suggest a change to
GNOME-PackageKit's GUI frontend.

(First of all, I'm still running Fedora 9, so perhaps in F10 and the
new release this has been remedied).

It is a significant aggravation to be unable to browse package lists
using the Add/Remove software client whenever the packagekit daemon is
already in the process of doing a system update, or already fetching
lists for another task.

And even when I have the daemon's full attention, the fact that it has
to reload package information straight from the repository every time
I select a new item is very annoying, especially since I may have just
selected that same item a mere second ago.  It is doubling irritating
when, due to internet connection irregularities, some items seem to
load immediately, while others may take several seconds.

Is there any way that packagekit could cache its own small database of
package names and descriptions, thus removing the need to download
fresh data for every single choice I make?

I am aware of course that this is merely correcting a possible design
flaw in Yum (that is, the flaw that Yum does not allow access to the
package database whenever in the middle of another task).  But since
Yum mainly just retrieves the information and handles the
transactions, I think it would make sense if the backend that ran Yum
then aggregated all the retrieved information on its own.

It would be handy for, example, browsing the repository when not
actually connected to the internet (especially if this could be
combined with Yum's package cache feature, so that in fact, complete
package browsing and reinstallation for recent packages could all be
done offline, or at least with a minimum of duplicated network
activity.)

I'm not sure how well this would map to other PackageKit backends, but
it would improve the experience greatly with Yum and Fedora.