Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.csd.uwm.edu!bionet!agate!ucbvax!hoptoad!gnu
From: gnu@hoptoad.uucp (John Gilmore)
Newsgroups: news.software.b
Subject: Re: "Lines:"
Message-ID: <8452@hoptoad.uucp>
Date: 6 Sep 89 09:48:23 GMT
References: <5200@looking.on.ca> <1989Aug30.052459.1166@vicom.com> <1989Sep5.145752.16320@mcmi.uucp>
Organization: Grasshopper Group in San Francisco
Lines: 34

It's trivial for each site to recompute the Lines: value if it wants
it.  NNTP servers can compute it for their clients, either when they
store the message, or when the client asks for it.  If a news reading
program doesn't want to read way out to the end of a 60K message before
displaying the first screenful, fine -- it can read a few hundred lines
and report the message as "long" rather than give an exact count.
I mean, jeez!

There is no point in transmitting it around the countryside.  My uucp
summary for August shows that hoptoad received more than 170 megabytes
and sent almost 300 megs, spending almost thirteen full days (319
hours) on the phone.  Thank you Telebit! but I wouldn't object to
cutting out 1% of that by dumping Lines: headers!  I fully agree with
Geoff and Henry that "Lines:" is one of the bogus headers that is
polluting the Precious Bodily Fluids of our network.  (One way that we
disagree is that I think wasting bytes on uselessly long message-ID's
is also a problem; they don't.)

On the other hand, a good cryptographically secure hash function to
provide a checksum of messages would be quite useful.  Usenet has been
dropping and mangling messages for too many years now; we can do better
with minimal effort.  I envison the checksum being in the message-ID so
that if you see a reference to a message-ID and later obtain the
message, you are guaranteed that you actually got a true copy of the
message in question.

(A cryptographically secure hash function is one where it is quite hard
to change the message and predict what result that will have on the
hash value.  For example, changing the meaning of the message while
leaving the hash the same is hard in a cryptographically secure hash
function.  It's not hard at all with things like CRC's.)
-- 
John Gilmore      {sun,pacbell,uunet,pyramid}!hoptoad!gnu      gnu@toad.com
		"Watch me change my world..." -- Liquid Theatre