[Yum-devel] [PATCH] Adds the new yum-cron.py and yum-cron.conf files.

[Zdenek Pavlas]

> Rather than try to modify yum-cron as it is, I
> think it will be easier and more maintainable to rebuild it more or
> less from scratch.  I've begun doing so, by heavily modifying the
> code from yum-updatesd to remove the code to run as a daemon and the
> dbus integration, and to make it easy to run as a cron job.

Hi, I think this is a good starting point, also gives you
the extra flexibility over running /usr/bin/yum directly.

> I'm interested in
> what people think of this approach, and what features and
> functionality should or shouldn't be added.

Ok, I have some ideas:

1) Finer locking

"download_updates = yes && install_updates = no":  This case
might deserve special treatment, with some care it might
be possible to drop the lock before downloading updates.

2) Conditional populateUpdates()

Searching for updates is IO and CPU intensive.  In many/most
cases both repo metadata and rpmdb didn't change, so we're
just calculating (and reporting) the same set of updates again,
which is pointless.

Maybe it's better to report each "update available" only
once?  That's extra state info, but OTOH we could exit early 
if metadata and rpmdb didn't change since our last run.