[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Automated package-popularity survey (was Re: We need easier installation.)

On Sat, Oct 24, 1998 at 10:51:32AM -0500, john@dhh.gt.org wrote:

> Matthew writes:
> > Pico probably beats xeamcs 'cos a lot of my users like it.
> I use emacs almost costantly, but I start up two copies when I start X and
> just leave them running.  As I restart X about once a week, emacs would
> probably come in way down my list despite being my #1 application.

At least the top 50% of the list on any one computer probably qualifies as
"popular" to that person.  I mean, the top 10 packages isn't that useful for
statistical purposes, because textutils, shellutils, grep, and findutils
will always be way up there.

If you've used a package in the last week or two, you probably use it. 
There are a lot of heuristics going on here, unfortunately... but I think
with the data from a lot of people, it can be combined into something
useful.  For example, if emacs were on my computer, it would never have been
run.  That's _very_ different from having run it last week.  (I'm a joe/xjed
split-brain patient).

Perhaps we can say:
	if it was accessed more than 24 hours after its attribute change
	AND it was accessed less then 30 days ago
	then it gets one vote.

My immediate response to that is it biases against packages which are
upgraded often.  Any other ideas?

Have fun,


Reply to: