innd and rejects of dates

[Julien ÉLIE]

Hi Ian,

> Russ Allbery writes ("Re: innd and rejects of dates"):
>> Nothing.  Injection-Date should be signed when issuing control messages as
>> well, but that gets tricky since the posting agent isn't supposed to
>> provide one.  I suspect what that means is that control message processing
>> software should impose a Date cutoff regardless of Injection-Date.
>
> Control message processing software doesn't know the server's history
> expiry date, or Date acceptance threshold, directly.  So that would
> would be yet another knob to have set incorrectly ...

Well, if you speak about controlchan, it has access to $INN::Config::artcutoff
so the problem is solved.  I am still unsure about what to do when the
value is 0 (= infinity), that is to say whether we should put an upper limit.
Anyway, when the value is 0, the previous behaviour of INN was to accept the
control article, and therefore process it...  So maybe we should do the same.

The only side effect I see is that controlchan would now also require
the DateTime::Format::Mail Perl module (so as to parse the Date: header field).
The check can be done at configure time.

(For Debian users, it is the libdatetime-perl package.)

Besides, I also believe an option should be given to controlchan, in order
to bypass the check on date (as a matter of fact, one may want to run
controlchan on an article -- and it would refuse to process it, if too old).




And if you speak about other scripts that can be used to handle control
messages, then they should try to read artcutoff...  It can be done easily
with Tcl, Perl and shell (via innshellvars).  Otherwise, innconfval could
be of help.

-- 
Julien ÉLIE

« Si l'art n'a pas de patrie, les artistes en ont une. » (Camille Saint-Saëns)