RFS Processing Analysis
I've been looking at the RFS process [1]. Some of this was in the
debian-mentors list about a year ago, albeit using corrupt data.
First the good news:
- About 1000 RFS submissions have led to accepted packages since January 2012
- The package acceptance rate has consistently been about 2/3
- About 1/3 of acceptances occur within a week [2][3]
The challenge - the RFS user experience degrades quickly after the
first couple of days:
- 1/2 of all active RFS submissions have no responses, despite a
median age of 2 months [4]
- 20% of all RFS acceptances happen after 3 months, often much longer than that
- For all submissions, sponsorship-requests averages 1 response per
RFS-month [5]
Also:
- For 'rejected' submissions, the first response is typically a drop
message with respect to mentors.d.n, at 20 weeks out
- The submission rate has been increasing at 40-50% per year
The interesting insight is that added monitoring of the debian-mentors
mailing list is the wrong way to address the deficiencies - the
packages most in need of attention are invisible there. The best way
to improve the average user RFS experience is to bypass the list and
target neglected submissions more directly.
That is not as easy as it could be. The mentors.d.n site shows plenty
of information about the backlog, but the main page only goes back a
week. The BTS summary shows all open submissions, but sorted somewhat
oddly, and showing only the package name and version.
Better tools may help. As a proof-of-concept, see the RFS Discovery
page [6], which exposes and sorts by comment activity. I find that a
better means of finding effective places to respond.
mentors.d.n could be a good home for such tools. Should that site,
with it's rich structured RFS data representation, grow to subsume BTS
as the canonical data source for the RFS process?
I used to have a story about one particular active RFS. The submitter
never failed to resolve substantive raised issues on the same day,
despite delays between comments of weeks, months, and even more than a
year. That RFS was recently closed to 'clean the queue'. I suspect
that user's opinion of the Debian development process could be
improved.
1) http://davesteele.github.io/debian-rfs-stats/
2) e.g. July 2014 - see 'accepted Packages'
http://davesteele.github.io/debian-rfs-stats/getslice.html?start=1404172800000&end=1406851200000&state=open#accepted
3) Where 'acceptance' means that the version submitted, or newer, is
present in sid or experimental.
4) http://davesteele.github.io/debian-rfs-stats/commentactivity.html
5) http://davesteele.github.io/debian-rfs-stats/mdbr.html
6) http://davesteele.github.io/debian-rfs-stats/discover.html?field=commentage&order=a
--
"Le mieux est l'ennemi du bien" - Voltaire
Reply to: