Re: [RFC] New task: science-dataacquisition
On Thu, 11 Dec 2008, Jordan Mantha wrote:
...
I think that is largely the state of afairs. Morten has done excellent
work and I've shifted essentially all my scientific packaging pursuits
to Debian Science and Debchem. I even became a DM to help do as much
work as I'm able to in Debian with the limited time I have. Part of
the issue has been, I think, that Ubuntu is primarily IRC-driven and
Debian is much more mailing list-driven so there sort of a
communication mismatch there. I don't think it's an unsurmountable
hurdle or anything, it just needs to be taken into account.
Ahh, OK. Thanks for your work and this explanation.
I certainly wouldn't put you in that corner :-) Ubuntu needs to be
able to take constructive criticism, just like everybody else. For my
part, as the guy that started the MOTU Science team and then had to
let it go as my dissertation got started, I think there is distinct
room for improvement in Ubuntu's handling of science packages that I
just haven't been able to address. For that I do feel bad. If Debian
Science has some suggestions and/or tips that might help things go
better I think Ubuntu would like to know.
As I suggested I think it would make sense to join here on the Debian
mailing list. IMHO it is the vital interest of Ubuntu science team to
have Debian science packages in good shape because it costs less work
in the end.
We use a Debian/Ubuntu version tracker [0] to give us information on
how we're doing. We are tracking 696 total packages.
IMHO it would make perfectly sense to compare your list with our tasks
files [2] and see what might be missing (on both sides). The problem
I have with the list [0] is that just from having the names listed
without any descrioption I fail to see the relevance for which specific
science.
IMHO it would be a very reasonable thing to do to work down the list[0]
and do something like this:
for pkg in "Packages from Debian" ; do
if ! grep $pkg science/trunk/debian-science/tasks/* ; then
apt-cache show $pkg
echo "Depends: $pkg" >> <tasks file which fits>
fi
done
With this simple algorithm we would make sure that Debian is making
profit from the work of the Ubuntu science people who did obviosely
some research on the package pool.
BTW, what do you think about our metapackage technique and the web pages
we are rendering out of the tasks files? IMHO it might make sense to
have a look at the bugs pages [3] as well - if the quality of the Debian
package is good you will not have to deal with much bugs in Ubuntu.
Of those only 5
packages ( 0.7%) are not in Sid. Those 5 are old packages we've
imported from elsewhere and frankly aren't worth Debian's time.
So you are talking about the "Not in Sid : 5 packages" of [0]?
I think Morten is working on some of them and for some purpose
they might be worth the work.
We
have 4 packages that have a newer version in Ubuntu than what's in
Sid. We should certainly be getting those back to Debian where
possible.
The "Outdated in Sid" packages of [0]? Did you filed any
"New version available" bugs? I hope that the situation becomes
better once I managed to build watch pages (like I did for the bug
pages[3]).
Right around 90% of the packages are taken from Debian
without any modification. Only 1.5 % are out of date in Jaunty
(current devel release) with respect to Sid. Overall I think we're
doing a pretty good job of sticking close to Debian where we can and
not lagging.
Could you please give some insight about this build process? It is
done somehow automatically or it is just like Ubuntu maintainer downloads
Debian source package, modifies changelog, builds and uploads?
The biggest problem I think Ubuntu is having is getting overloaded
with bugs. We have 365 science bugs open currently [1] and that's just
too much for a handful of people to try to get all triaged and
forwarded quickly.
How are these bugs related to the Debian BTS status? Are there a lot
of bugs concerning the same issues? Are the bugs forewarded to Debian
BTS if it is unknown here?
No flamewars needed :-) I think Debian and Ubuntu have a lot more in
common than not.
Thanks. I would really love if discussion culture on debian-science
list is better than on debian-devel ...
Kind regards
Andreas.
[0] http://qa.ubuntuwire.com/multidistrotools/science.html
[1] https://bugs.launchpad.net/~motuscience/+packagebugs
[2] http://svn.debian.org/wsvn/cdd/projects/science/trunk/debian-science/tasks/?rev=0&sc=0
[3] http://cdd.alioth.debian.org/science/bugs/index.html
--
http://fam-tille.de
Reply to: