Re: [9fans] size of Plan 9 user base

[Richard Uhtenwoldt <ru@river.org>]

>Can you also provide stats for the number of messages,  so we can
>compare ourselves to the *BSDs?

here you go.  the archive for the FreeBSD lists does not list
folder sizes, so no i have no info on BSD.

but i subscribe to lists covering the Erlang and Haskell
functional-programming languages, so here's a
summary of traffic volume from 2002 dec 1 to 2003 may 1.

  plan9    9.9 megs in 1242 messages,
  haskell 28.5 megs in 781  messages,
  erlang  13.0 megs in ???  messages,

where
plan9 == 9fans
erlang == Erlang-Questions <erlang-questions@erlang.org>
haskell == haskell@haskell.org, haskell-cafe@haskell.org,
    glasgow-haskell-bugs@haskell.org, glasgow-haskell-users@haskell.org,
    ffi@haskell.org, nhc-bugs@haskell.org

the data above imply that on average, a haskell message is at
least 3 times the size of a Plan-9 message!   so, a check
of the data is in order.

for the check, I did not take pains to start and stop at specific dates:
rather I just picked 3 files of plan9 archives and 3 files of haskell
archives.

since we are interested in a ratio (average characters per message), we
do not need to ensure that one set of archives covers a different start
(or end) date.

$ wc -c plan9.old[def]
5089705 plan9.oldd
5701148 plan9.olde
5568847 plan9.oldf
16359700 total
$ grep 'Return-path: <9fans-admin@cse.psu.edu>' plan9.old[def]|wc -l
   4478
16359700 divided by 4478 messages
   == 3653 characters per messsage

$ wc -c haskell.old[bcd]
4751391 haskell.oldb
3023976 haskell.oldc
3672447 haskell.oldd
11447814 total
$ grep 'Delivered-To: haskell@haskell.org' haskell.old[bcd]|wc -l
    808
11447814 divided by 808
   == 14168 characters per message.

$ wc -c erlang-
24533390 erlang-
$ grep 'Return-path: <owner-erlang-questions@erlang.org>' erlang-|wc -l
   5150
24533390 divided by 5150
   == 4763 characters per message.

summarizing,
the avg 9fans   message is  3653 characters.
the avg erlang  message is  4763
the avg haskell message is 14168

so, on average, a Haskell message really is bigger:
3.8 times the size of a 9fans message.

the haskell lists apply to every message a "footer" that says, "to
unsubscribe to this mailing list...", etc, of 137 characters.

but that 137 bytes is trivial compared to the differences
in average message size (over 10,000 bytes).

(i'm relying here on the fact that subtracting 137 bytes
from _every_ message causes the _average_ bytes per message
to decrease by exactly 137 bytes.)

finally, I tried to explain the discrepancy by measuring the size of the
headers of a few messages, but that does not explain it either.

about half of the Haskell posters are professors, and I suppose
professors are just more verbose.