From: Charles Lindsey (chl@clerew.man.ac.uk)
Date: Thu Jul 08 2004 - 15:06:28 CDT
In <87658zp561.fsf@windlord.stanford.edu> Russ Allbery <rra@stanford.edu> writes:
>Charles Lindsey <chl@clerew.man.ac.uk> writes:
>> 1. I note that son-of-1036 did include
>> "For historical reasons, the subject MUST not begin with "cmsg " (note
>> that this sequence ends with a blank)."
>> followed by a NOTE explaining why and declaring the cmsg-hack as
>> "obsolete". That is probably the reason why various agents currently
>> reject such Subjects, as noted by Frank. It would be useful to know how
>> many current agents actually do that. Does INN, for example?
>I like that statement; it's nice and simple.
>INN (nnrpd, to be precise) adds a Control header matching the contents of
>the Subject header following "cmsg ", provided that no Control header is
>already present. From what you mention later in your message, I'm
>guessing that tactic was taken straight from C News.
Let me be clear.
1. In current versions of INN, the injector (POST command/whatever)
modifies the article by adding a Control header (so now every site on
Usenet will see it and obey it). That is way more than RFC 1036 asked for,
and I think we are agreed it should stop.
2. In earlier versions of INN, it also did that but _in addition_ it also
caught incoming (relayed) messages with cmsg (but no Control header) and
treated them (internally) as control messages. That is (more or less) what
RFC 1036 asks for.
3. There is a proposal (in son-of-1036 and in our earlier drafts) that
injectors MUST/SHOULD reject articles whose Subject started with "cmsg"
(and did not contain a proper Control-header as well).
Clearly, #1 and #3 cannot coexist on the same injector. #1 and #2 can
coexist on the same server. #2 and #3 could in theory coexist, but would
be bizarre.
I assume we intend to forbid #1 outright, so it either gets caught by #3
(if #3 implemented), or goes out as a normal article (if #3 not
implemented).
>That's obviously not the best behavior. I would be interested in what
>this group would recommend instead; I'm happy to implement the suggestion
>of the group in the next version. Should such messages simply be rejected
>if no Control header is present? (Note that there is no way in the NNTP
>protocol to issue a warning about a message but post it anyway.)
First abolish the behaviour (#1) and then reject (#3) if that's what we
decide to say in the draft.
>> 2. Our draft contained a similar sentence at one stage, but it got moved
>> to USEAGE (and demoted to SHOULD) as part of the great split. Nobody
>> commented on that at the time. Replacing that text in USEFOR is clearly
>> one of the options open to us.
>I think we should restore that statement. I'm ambivalent as to whether it
>should say MUST NOT or SHOULD NOT.
Let us look at the tradeoffs:
There exist, and will continue to exist, servers which
a) implement #2 (bad)
b) implement #3 (to protect against #2)
c) do neither (treat as normal articles)
It would be useful have some idea of how many of each kind there are but
suppose, for the sake of argument, that there are 33.3% of each.
Assume our draft deprecates #2. Then the number implementing #2 will
wither away as sites become compliant.
Two possibilities remain:
p1) our draft encourages #3 (as you just suggested). The number
implementing #3 increases as sites become compliant, and at roughly the
same rate as the number implementing #2 goes down.
p2) our draft does not encourage #3. The number implementing #3 goes down
as sites become compliant (but possibly more slowly than with #2).
The final steady state is:
Under p1) No sites implement #2. All sites implement #3, which has now
become pointless :-( .
Under p2) No sites implement either #2 or #3, which is the ideal final
state. "cmsg" is finally dead.
Now one could do some algebra to work out the intermediate states, under
various assumptions of proportions of sites of types a), b) and c) and
some assumption as to how often people actually try these tricks
currently. But it is not clear to me which of p1 and p2 would come out
best. But I will let someone else do the algebra, because I am away from
tomorrow.
But one should also remember that we already have sites of types a), b)
and c) (in some unknown proportions), and so one could calibrate the
exercise by asking "how much grief is caused by the problem in the present
Usenet". The less that is, the more appealing p2 would appear.
>> 3. Alternatively, we insert a warning as suggested above. Here is a
>> possible text, to replace the present NOTE in 6.13
Yes, that clearly need reworking in the light of comments.
-- Charles H. Lindsey ---------At Home, doing my own thing------------------------ Tel: +44 161 436 6131 Fax: +44 161 436 6133 Web: http://www.cs.man.ac.uk/~chl Email: chl@clerew.man.ac.uk Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K. PGP: 2C15F1A9 Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5