Re: Weekly posting summary for [e-mail]

[John C Klensin <[e-mail]>]

--On Friday, 07 June, 2013 10:57 -0700 Bob Hinden
<[e-mail]> wrote:

> Thomas,
> 
>> From my perspective, the intention/usefulness of the weekly
>> posting is to give folk a high-level view of who is posting
>> and how often. It is not uncommon to see certain individuals
>> stand out. In some cases, that makes perfect sense -- and the
>...
> I agree with your conclusion.  
> 
> I also wish that the signal level was higher.  The IETF list
> is important because we need to have a place where IETF wide
> issues can be discussed.  That doesn't work if many people
> unsubscribe due to the the low signal to noise level.

Agreed so far.

>  I hope
> that that people who consistently show up at the top of the
> posting summary will moderate their behavior.

And it is getting to that conclusion from the above that often
troubles me about the posting summary list rankings.  Assuming a
significant issues shows up on the list, whether in conjunction
with a Last Call or something else.  Posting a comment and then
following up the comments of others with a couple of more
postings constitutes three messages in a week, which is pretty
reasonable.  On the other hand, if there are four such issues in
a single week (it happens) then that same individual gets
"credited" with a dozen messages, which would make the top of
the list in many weeks.  

Because things do often seem to happen in clusters, the summary
list might be improved significantly by adding a four-week
average and rankings based on it.  That would help distinguish
the regular heavy posters from those whose numbers went up
because of a particular issue or set of issues in a given week
or two.

What that still wouldn't help with is distinguishing from a
theoretical very active and diligent IETF participant who
reviews and comments on every document that goes into Last Call
push every other issue discussed on the list and a troll who
comments on every Last Call and on other issues that he, she, or
it might even generate.  To measure that difference, one would
need a measure of the useful and relevant information content of
a message, not just a character/byte count.

best,
   john