Re: [ICANN-EU] MEMO requesting discussion: Structuring a large ML

["Constantine S. Chassapis" <cschassapis@acm.org>]

Hi all,

I know this has nothing to do with UDRP which is the hot topic
for now, but, some messages ago (maybe also some days ago),
I said that I will perform a simulation to establish if my pointing
system has any future. 

I thus made a probabilistic simulation (in fortran 90) of an ML 
of 15000 members who post on the average 1500
messages per clock tick (lets say per day) and where 420
of them get threaded per tick, and where 200 SUMUPS were done
and where about 60 were not done. The members were divided
in five classes (tourist - 5%, lazy - 30%, average - 40%, 
hardWorker - 20%, exceptional - 5%), and their responsiveness
or eagerness was varied. Also their eagerness varied 
according to the points they secured (it dropped). The 
methodology was a sigmoid function that set the probability of
response and its center varied according to class and points
collected.

I run many simulations for 10000 time-steps and found that my
orignal pointing system failed due to the exponential growth of
links (from the collectors).  So, I found that the following
pointing system had a stable asymptote (after 100000 steps): 
4165 collectors, 10272 simple members and 563 temporarily 
frozen members:

+100 at entry.
+0 per message posted.
+1 per SUM UP done.
-10 per SUM UP not done but required.
-1 per improper message.
+1 per message that starts a thread or is linked back in a
   collector's collection.
+1 per message that is linked back in a collector's 
   collection.
 +1 per collection to a collector
the limit for denial of posting ability was set to -100 
the limit for becoming a collector to +300

(Final note: The asymptotic number of collections was 172 per step,
and these collections had 1500 links per step all together. The
percentage of improper messages was set to 10%).

And that completes the story. I will not add anything more
here unless somebody asks so. The system is there in the
MEMO and its supplements. If anyone wishes to see the
code, I will be happy to give it.

Maybe the system proposed by others, of two or three parallel
MLs, is better. In any case, we should self-organize fast.

Sincerely,
Constantine S. Chassapis

P.S. -- If I will ever publish something in the form of a paper 
about this stuff, you (the 15 or 20 people that have posted your
ideas in this list) will be the first to be informed.

P.S.2-- Hey Thomas, it is really interesting that only 15 to 25
people post messages in the icann-europe@fitug.de. Do you
know how many members this list has?