On Tue, 2011-12-13 at 13:28 -0500, Genes MailLists wrote:
On 12/13/2011 09:21 AM, mike cloaked wrote:
>
> Indeed so - however it is not unusual for abrt to require pulling in
> several hundred megabytes of debuginfo stuff in order to file a report
> - and I wonder if that might put a significant number of users off and
> abandon completing the report at that point, particularly if they have
> limited bandwidth available for downloads?
>
> This may already be inhibiting the provision of some fraction of
> important abrt entries? It would be interesting to hear other people's
> view on this aspect?
In my opinion abrtd should not download any debuginfo's on the users
machine - zero. As long as the developers have the debuginfo the
tracebacks are useful - so there's no point in forcing users (devs can
download whatever they need) to download any debugging.
That said, Alan Cox mentioned in a different thread that there is a
version of abrtd which sends the data to a server (which has all
debuginfo files) to analyze - that is better in my view.
I tried this a while back. All I got was a "server overload" error, and
IIRC that took a long time to show up. In any case, it didn't work.
I tend to agree that this would be better done on the server side, but
someone needs to do some load calculations. Maybe server-side for the
base packages and client-side for the rest or something of that sort.
poc