enhance checkgroups handling
rra at stanford.edu
Mon Aug 13 17:19:38 UTC 2007
Julien ÉLIE <julien at trigofacile.com> writes:
> I agree with you. I tend to think it would be better to honour such
> "doit" actions. I asked this because of the default control.ctl
> For instance:
> ## AR (Argentina)
> checkgroups:jorge_f at nodens.fisica.unlp.edu.ar:ar.*:doit
> newgroup:jorge_f at nodens.fisica.unlp.edu.ar:ar.*:doit
> rmgroup:jorge_f at nodens.fisica.unlp.edu.ar:ar.*:doit
> Everyone can make a mess with checkgroups (2000 ar.* newsgroups created
> in the row with a forged checkgroups... and then another checkgroups
> which deletes them and afterwards another which...).
> But well, this can currently also be done with newgroup/rmgroup articles
> (but the mess is softer).
Maybe we should put some sort of limit on how many changes can be made
with one checkgroups without the administrator manually running
docheckgroups with some flag? Sort of like what actsync does.
Also, make sure that the checkgroups handling in INN actually obeys the
pattern in control.ctl. It used to have problems in that area, but I
haven't looked at it again since your recent modifications.
> I do not know what to do with that. It can lead to possible attacks
> (and on ftp.isc.org too by the way).
Yeah, I've been not worrying about that until someone actually did it
since creating newsgroups doesn't consume any actual resources on
ftp.isc.org and it's easy to reverse.
> Wouldn't it be time to only allow PGP-signed control articles to be
> honoured? (inviting current hierarchies maintainers to sign their
> control articles) Active hierarchies are mostly PGP-managed, though.
> [perhaps we should discuss that in news.admin.hierarchies]
True. Maybe it's time to take that step.
Russ Allbery (rra at stanford.edu) <http://www.eyrie.org/~eagle/>
Please send questions to the list rather than mailing me directly.
<http://www.eyrie.org/~eagle/faqs/questions.html> explains why.
More information about the inn-workers