From: greg andruk (meowing@banet.net)
Date: Fri Nov 05 1999 - 01:06:25 CST
On 4 Nov 1999, Russ Allbery wrote:
> INN is very, very picky about duplicated headers. It will reject any
> article that has a duplicated Control header, both local posting and
> transit
Not quite. Duplicate controls aren't allowed at posting time [that's all
relatively recent stuff, along with the automagical cmsg... conversion and
the hardcoeded list of known control messages], but they're okay-fine for
transit. Just tried it a moment ago, that's still true in the current
code.
[nnrpd has lots of policy stuff hardcoded in C which is more appropriate
for local filters, but that's another story.]
parsecontrol and controlchan have either ignored additional control
headers or declined to process articles with more than one control
header, because historically they've been mostly unintentional or sent
with ill intent.
It certainly wouldn't be that big a deal to have these loop through the
control headers and decide what to do on each one, but before that would
be reasonable there are permission problems that need to be worked out
(should mixes of allowed and disallowed newgroups and rmgroups be
partially processed or thrown out altogether?).
ALso, the sensible combinations need more clarification. Would there be
there a good reason to allow, say, a combination of newgroup and
checkgroups? Which, if any, parts get honored?
> (as well as a duplicated just-about-anything-else-it-knows-about).
> I intend to fix this, but it requires redoing the header parsing code and
> I haven't had a chance yet.
Only the required headers get checked for duplicates in transit, which in
earlier sections seems to still be what we want. nnrpd just needs some
comparisons (after the main parsing) yanked out.
On that note, INN still goes to the trouble of stripping some ancient
headers in transit, or rejecting articles with them at posting time. These
headers are Received:, Date-Received:, Posting-Version:, Relay-Version:
and Posted:. Is there any reason to continue caring about these, and does
the standard need to mention them?