Michael DeHaan napisal(a):
> * for the entire membership, which domains are they coming from?
> * for these categories of membership (core, active, passive) which
> domains are they coming from?
>
I don't find this categorization particularly meaningful. It starts
to get to the point of trying to classify someone
into groups that really don't exist.
Just because there's no formal structure that doesn't mean there
is no structure at all.
What's going on on mailing lists? People are trying to lobby their
ideas/patches into the repository. How do they do it? By mailing. The
more mails the better lobbing. Who do they have to mail to? Core
developers. You won't see your patches applied if you're mailing
with temporary users. You have to point your guns at core developers.
The more mails sent to core developers the better lobbing.
The core developers have to speak to core developers to agree
on some solutions, but they also have to mail to the lobbysts
and tell them 'no way'.
And there is a structure behind it. It is a structure of power.
Two the most powerful groups are: core developers just because
they have +w right on the repository and the regular users,
because they may vote for changes, and they are the real lobby
force. The rest is the temporary users.
The "talking to" algorithm would be error prone on mailing
lists.
Folks reply to questions
more so than who posts them.
We clearly would not want to be developing fuzzy logic to determine the
type of question.
You are overestimating my project :) The algorithm does not care if the
mail sent is a question, a link suggestion, a patch, a mistake or
something else.
The thing is that, it tries to find some patterns in large sets
of data structured by the reply mechanism. It's based on statistics,
so it may make mistakes. But all in all it should point you to
the mailing lists and the projects which may need some kind of action
(just as I described it in my previous mail) or which are at the
interesting point of development.
If the goals of EKG project are much different than what I described
then there is really no point in implementing it. The only problem is
that at the moment I can't see clear strategy of future development,
and outsiders like me may find it confusing. I mean, the ideas page says
that GSoCers should come with their ideas of data mining,
but it's really hard to guess what else can we mine in here if we
do not agree to do something more complicated than counting the domains.
I do not feel very attached to my initial idea. It may go to the bin.
But I think that there is not enough pointers for GSoCers to go the
way you would like them to go.
Cheers,
--
Regards,
Wojtek Walczak,
http://tosh.pl/gminick/