[Ubuntu] Fwd: [GRASS-dev] ubuntu gis policy

[Angelos Tzotsos]

On 09/16/2016 05:29 PM, Sebastiaan Couwenberg wrote:
> On 09/16/2016 03:45 PM, Micha Silver wrote:
>> Randal Hale
>>> I'm one of the guys who was commenting on the confusion. Someone
>>> commented that the naming is in keeping with Ubuntu policy. Which
>>> probably for the Sys Admin crowd is perfect. We've got a weird blend
>>> of Users who want software BUT (and I didn't up until this last
>>> thread) understand the reasoning behind the naming. It is confusing. I
>>> don't know if something on the PPA explaining it is warranted OR a
>>> name change OR something on the package websites.
>>>
>> I'd suggest this: (clearest to me, and does not break the known Debian
>> chain)
>>
>> "Experimental" (=Debian testing) --> "Current" (=Debian unstable) -->
>> "Legacy" (=Debian stable)
> The above is wrong. Packages in Debian migrate from unstable to testing
> after a period of 5 to 10 days (depending on urgency of the package, 5
> days for medium, 10 days for low) assuming no release critical bugs are
> affecting the version of the package in unstable and all of the required
> versions of dependencies are in testing or will migrate to testing on
> the same day as the package in question.
>
> Debian testing is therefore less prone to issue than unstable in which
> active development is done. The experimental repository is used for
> packages which should migrate to testing or are otherwise unfit for
> inclusion in the next Debian stable release via the testing repository.
> This includes packages which require a transition (rebuilds of all its
> reverse dependencies) to have a coherent package set migrate to testing.
> Not all maintainers adhere to this, and uncoordinated transitions are
> triggered by uploads to unstable which should have gone to experimental
> first. We've recently had this for libdap & openmpi for example.
>
> The flow of packages in Debian is as follows:
>
>   (experimental ->) unstable -> testing -> stable
>
> New packages are uploaded to unstable by default, or experimental if
> more work is required before they can migrate to testing (e.g. a
> transition or wait for the final release in case of beta & RC
> pre-releases), after the grace period the packages from unstable migrate
> to testing, packages in testing get frozen eventually in preparation of
> the next stable release. Because codenames are used for the
> distributions the packages in testing aren't copied to stable, the
> stable suite is simple renamed to oldstable and the testing suite to
> stable. Hence the importance to use release codenames in apt
> sources.list to stick to your chosen distribution and be forced to
> switch when the new stable release is out.
>
> Because of the abuse of the ubuntugis-testing repository for
> experimental packages, that repository should be renamed to reflect its
> use. Users and administrators of systems using the UbuntuGIS PPAs who
> are uncomfortable by unpredictable changes should stick to the stable
> PPA which is only infrequently updated. Reintroducing a testing PPA in
> which more testing is performed by the users before those packages are
> moved to the stable is valuable to root out bugs and integration issues,
> but the UbuntuGIS userbase is not as involved as in Debian so the amount
> of testing those packages would receive is likely to be very limited if
> not non-existent. The packages in the UbuntuGIS PPAs are also not
> subjected to automated QA as the packages in Debian are. Making the
> value of an UbuntuGIS testing PPA questionable in practice.
>
> Most UbuntuGIS users use the unstable PPA because they want newer
> upstream versions than the packages included in the Ubuntu LTS release,
> and are not willing to wait until those version become available in the
> UbuntuGIS stable PPA.
>
> Until more development manpower is available, especially someone willing
> to lead the integration & testing work required for stable releases,
> discussing the PPA naming is of little to no value. All eyes are focused
> on the unstable PPA with hardly anyone caring for the other PPAs. If it
> weren't for the updates from OSGeo-Live the UbuntuGIS PPAs would stop
> being used due to irrelevance. Having lots of users doesn't mean having
> enough developers to match. This is common in the Ubuntu world which has
> lots of users but much less developers when compared to Debian. Ubuntu
> users should learn to become developers for the ecosystem to flourish.
>
> Kind Regards,
>
> Bas
>

Thank you Bas for the detailed description.

There have been many cases in the past, where system admins from well 
known organizations have said: "we accept packages from Debian/Ubuntu 
upstream OR UbuntuGIS". Some of them have explicitly asked for UbuntuGIS 
stable or else they would not accept to deploy packages from other 
sources. This IMO means that there is a developed trust of people to our 
stable ppa even if the packages there are not changing often. But as 
already said, our unstable ppa is still the most popular, due to the 
update rate.

I believe that we should not touch the names of stable and unstable ppas.

Now regarding testing ppa, I think it was dead before we decided to sync 
it with OSGeoLive nightly, so renaming it to experimental makes sense to 
me. I agree that there is no man power available at the moment for an 
actual Testing ppa, I think the OSGeo-Live releases practically have 
this role today. This is why I have proposed to move well tested 
OSGeo-Live packages to stable after some time.

Best,
Angelos


-- 
Angelos Tzotsos, PhD
OSGeo Charter Member
http://users.ntua.gr/tzotsos