We don't. The definition of the "core" is simple: More than 75% of the
average of replies sent and more than 75% of the average of connections
with other users. No more than that.
Ah, but you are doing this :)
Do you follow the point about telling the data what to say?
Why not 65.2%, or 86.3%? Etc. Such definitions imply added meaning when
there may not be one. A poster who
does not read the docs, and replies with "me too" could easily be
"core"
in your definition -- while one who is one of the most
important contributors in terms of testing, patches, or even marketing
could not post a lot.
These definitions are arbitrary, which is why we should not be grouping
users.
When you look at the results you should be aware of how things are
defined and I think that these definitions are simple enough to be
mentioned at the top of every report.
While I appreciate the interest, as part of GSOC it is important to
understand what we (Max, Yaakov, myself, etc) want to get out of the project
is what we're going to sponsor and you'd need to go along with that. We
might not want what all of the data you'd want to present/gather -- we'd
probably want different things.
You would need to bend a bit rather than saying this is the most
important thing and should be at the top of every report. Everything
can't go at the top :)
I also think that the way I called this group at the beginning is
bad.
It should rather be: core mailing-list members, regular and temporary.
Temporary might not be the best name too. I am not sure what's better
though, 'passive', 'low participation members'?
> Different people are great for different things.
>
> This is why EKG shouldn't try to assign meaning to the data -- it should
> be about gathering and presenting the data and let people draw their own
> interpretations from it.
>
Fully agreed. And I think that my proposition is not assigning more
meaning to the data than any other. What I wanted to present for end
users is:
mail-list name | exclusiveness level | responsiveness level | temporary
users participation level | pie of core members by domains | pie of
regular members by domains | pie of temporary members by domain
I don't think we like your group concept.
I think we'd be fine with tracking things in more statistical terms,
like the mean number of posts each user makes, the median time between
posts, standard deviations,
number of replies per thread, etc. Just don't try to interject meaning
into the data. Splitting the pies, for instance, would not be highly useful.
It is also important to make sure that the data is actionable -- that it
can imply something we (Max, etc) could do with it. Otherwise, we're
just showering
the user with a lot of noise, and it's easy to get lost in what is
important and what is not.
This is why visualizations are important.
I think in the course of the larger project -- some additional ML
statistics are good, but for GSOC we'd also like to see a bit more --
hopefully also covering
a bit on the visualization side and/or adding another data source on top
of just the above. However we don't want to be innundated with things
like graphs
of which lists get the most deeply nested replies (or some arbitrary
stat) when we could gauge the same findings from less. At least when looking
at the big picture, less is more.
Namely, from my experience from running projects, knowing how many
replies a thread gets is interesting, but it's not ALL that interesting
-- because I can likely deduce
whether the list is thriving or not just based on the post count. Some
of that data is marginally interesting in determining if a list is a
patch-only list that doesn't get a lot of review,
but the existance of such lists doesn't imply an unhealthy list either
-- it might just imply that something is not a users list, which we
probably already know (because we're humans
and we'll be looking at the list anyway).
Contributions come in many different forms. Some of your best users may
talk to you over IRC (which we may or may not track), or just send in
one huge patch every 5 months.
I wouldn't call these folks any more "temporary" or "core", and
such
distinctions aren't really useful. If we want to show the report of who
the top posters are for any given list, like version 2 did,
yes, we should do that. Drawing the line at "core" or "temporary"
though, no.
Hope that explains what we are trying to do a bit more.
--Michael