[packagekit] viability of the current yum backend ?

Robin Norwood rnorwood at redhat.com
Wed Jan 2 14:59:01 PST 2008

Richard Hughes <hughsient at gmail.com> writes:

> On Wed, 2008-01-02 at 17:16 -0500, David Zeuthen wrote:
>> Keep in mind that whenever you add caching you need to care about
>> coherency and whatnot. The devil in the detail here, broadly speaking,
>> is that cache coherency issues pops up when you least expect. And
>> normally it's very hard to debug such things.
> Totally. With the updates cache we drop it as soon as yum does anything
> behind our back, or for that matter, when we do anything crazy like
> manually updating the cache.
>> So unless it's worth it, try avoid caching as much as you can. Or,
>> rather, implement the cache as low in the stack as possible (e.g. in the
>> yum/rpm code base itself). This is because of the fact that the IPC
>> overhead from pk-update-icon<->message bus<->packagekitd<->yum-backend
>> is close to negligible compared to human response times (> 20ms)
> Point taken and processed. Robin was thinking about something clever
> with a python process but we'll have to bribe him with beer to get more
> information. :-)

Well, I don't know how clever it is, but, since the backend already
hangs around for awhile when actions are being performed, jbowes and I
were talking about just having the backend load yum once, instead of
each time a helper is spawned.


Robin Norwood
Red Hat, Inc.

"The Sage does nothing, yet nothing remains undone."
-Lao Tzu, Te Tao Ching

More information about the PackageKit mailing list