On Wed, Oct 1, 2008 at 8:43 AM, Barbie <barbie@missbarbell.co.uk> wrote:
> There has been a recent incident where an author had deleted a
> distribution from CPAN, but is still receiving test reports for them
> (via the Daily Report).
>
> I'm assuming that that tester in question is using a CPAN-Mini mirror or
> similar to retrieve the modules for testing. If this is the case for

Not necessarily.

Deleting something from CPAN is a multi-step process.  First PAUSE
schedules something for deletion and there is a 72 minimum delay
before the deletion actually takes place by a cron job.  Then there is
a propagation delay for all the mirrors to sync up when PAUSE actually
deletes the distribution from the master CPAN repository.  A
CPAN::Mini repository is no different than any other rsynced CPAN
mirror in that regard.

See http://www.cs.uu.nl/stats/mirmon/cpan.html to get a sense of just
how infrequently some mirrors sync up.  The median is 12 hours.

Ideally, CPAN Testers should use a frequently updating mirror to avoid
testing modules recently deleted from the master CPAN.  But that can't
be uniformly guaranteed.  And nothing prevents testers from testing
module that were recently uploaded and then immediately scheduled for
deletion in the 72 hour delay window.

Did the author get a report from a test run during the scheduled
deletion period or after it?

A better approach is probably to filter out or flag (for later
exclusion) any reports in the database for distributions no longer
available on CPAN.

-- David