Re: backports

[Yaroslav Halchenko <debian@onerussian.com>]

imho all sounds great and had been our approach for non-official
neuro.debian.net.

0.1cents in a form of aspects I think it might be worth keeping in
mind:

*  backports friendly packaging 

   do not overcome insufficiencies of *Depends versions
   specifications with strict limitations without necessity

*  convenient  (transparent) backporting

   straightforward tools nd_backport and nd_build4allnd [1] Michael wrote
   for our use were sufficient so far in most of the cases to provide
   backports for majority of Debian and Ubuntu releases

*  distribution specific patchsets

   sometimes transparent backporting is not feasable without patching,
   so enabling distribution specific patch series should be provided.
   See backport-dsc from the same repo
  
*  testing

   important fact, often disregarded -- Scientific software, as nothing
   else, is in need of unit- and regression testing.  Back- (or
   forward-porting for a distribution which might be slightly ahead,
   e.g. upcoming Ubuntu new release) ported software might not have
   higher chance to lead to incorrect behavior due to various factors...
   at least partially to prevent that -- testing during package building
   seems to be of great help and should be encouraged (if not enforced)
 
[1] http://git.debian.org/?p=pkg-exppsy/neurodebian.git;a=tree;f=tools;hb=HEAD

On Sun, 12 Sep 2010, Steffen Möller wrote:

> Hello,

> how do you feel towards the idea that with the advent of backports as an official service of the distribution, we should use that
> in routine for the more frequently updated tools? In science, there is rarely (i.e. except for comparisons or for continuity) the
> need to use an older version, and then snapshot.debian.org comes to a rescue. And then there are these annoying cases when a new
> upstream release just fails to miss the freeze.

> Prime candidates IMHO are the autodocktools, gromacs, ... well ... almost anything in debian-science and debian-med, really. I
> would then even opt to take the autodocktools out of the main distribution towards an appearance in backports and testing only.

> Comments welcome.

> Best,

> Steffen
-- 
                                  .-.
=------------------------------   /v\  ----------------------------=
Keep in touch                    // \\     (yoh@|www.)onerussian.com
Yaroslav Halchenko              /(   )\               ICQ#: 60653192
                   Linux User    ^^-^^    [175555]