Re: custom vs. derivative
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Wed, Apr 09, 2008 at 01:19:06PM +0200, Andreas Tille wrote:
> On Wed, 9 Apr 2008, Jonas Smedegaard wrote:
>> These fit your definition then:
>>
>> * A distribution made from all of main + 1 package from non-free
>
> Sounds reasonable to exclude non-free and specify more detailed what
> we want to include.
What sounds reasonable to me is to not invent any new rules for
exclusion or inclusion, but simply follow Debian. And explicitly state
so in our CDD definition.
See my updated proposal further down, and the ASCII art at the end.
Debian has serious reasons for its segmentation of packages.
When Debian consider something "non-free", then CDDs can choose to
separate even more (like grouping into _reasons_ for being non-free) but
may not circumvent the main separations.
When Debian consider something "testing", then CDDs cn choose to
subsegment that for their local needs, but may not promote it as stable.
This does not block CDDs from distributing a mix of stable, testing and
unstable packages, as they have all[1] evolved through the same earlier
distributions: A CDD containing stable packages except a single unstable
package must be promoted as unstable, and a mix of testing and stable
packages must be promoted as ither testing or unstable.
Or put another way: CDDs are restricted to same _distribution_ but not
same _release_ of that distribution: mixing across _time_ is ok[2].
[1] specially injected packages like stable updates have not gone
through the normal evolution, but I see no problem promoting those as
either testing or unstable even if they were never in those
distributions.
[2] To be really nitpicking, it is actually not ok to mix across time
for the _stable_ distribution: Debian does not support mixing stable
with stuff older than unstable, and it can cause real problems to do so,
as package maintainers are encouraged to clean up dependencies no longer
relevant.
Fundamentally, CDDs should be standing on the shoulders of Debian - not
hanging on its neck.
>> * A distribution made from all of stable + 1 package from testing
>
> Hmmm, thinking about ist: Why not? I did not had this in mind but I
> have no problems with this in case it is tested and just works. But I
> have probably not considered all consequences that might occure out of
> it.
See above.
>> A "Custom Debian Distribution" is s subpart of the "Debian
>> distribution" as defined by the Debian project. This means that all
>> parts of it must be completely contained inside the Debian
>> distribution.
I now understand that "Debian distribution" is not singular (each
official stable release is a distribution, and each daily release of
testing or unstable is a release), so here's a nitpicking corrected
proposal:
A "Custom Debian Distribution" is a subdistribution of a Debian
distribution. This means that all parts of it must be completely
contained inside a single Debian distribution.
See also ASCII art further down...
> I would like to you to put such things on the Wiki to make it
> converge to something useful there. Please work actively with
>
> http://wiki.debian.org/CDDNamingProposals
>
> because I'm afraid that good ideas will be lost in a lengthy
> thread on the mailing list.
Because you widened that page from naming CDD to naming other things
too, lengthening the discussion and blurring the topic IMO.
Feel free to use my input in that broader naming task, I just have no
interest in participating in that.
I see CDDs as _refinements_ to Debian itself without _distorting_ it.
* no blurring of what Debian defines as free
* no blurring of what Debian defines as stable (or testing)
* no blurring of what Debian defines as officially released together
The above is *only* examples. Not meant for inclusion in a definition.
I believe my proposed definition contains the above. If some of
you feel that you can read any of it into my proposed definition, then
please elaborate.
If some of you feel that CDDs _do_ contain any of the above, then please
elaborate too.
If, on the other hand, you feel that CDDs _should_ support any of the
above even if it does not currently, then I honestly don't care about
your input. Feel free to elaborate, I just won't participate down that
subthread.
>> This is important for me to clarify, as something mixing stable and
>> testing stuff is considered CDD development to me, but an official
>> release of a CDD cannot IMO be based on stuff that is not also
>> officially released by Debian.
>
> Just read
>
> http://people.debian.org/~tille/cdd/ch-todo.en.html#s-new_ways_of_distribution
>
> This occured out of some discussion at OSWC in Malaga 2004
Ah - thanks!
Here is my view on the above, illustrated like the referenced text:
unstable -> testing -> stable -> official stable release
| | |
| | +-> CDD_A stable -> CDD_A stable release
| | |
| | +-> CDD_B stable -> CDD_B stable release
| | |
| | +-> ...
| |
| +-> testing snapshot release
| |
| +-> CDD_A testing -> testing CDD_A snapshot release
| |
| +-> CDD_B testing -> testing CDD_B snapshot release
| |
| +-> ...
|
+ -> CDD_A unstable -> unstable CDD_A snapshot release
|
+ -> CDD_B unstable -> unstable CDD_B snapshot release
|
+-> ...
The drawing intends to reflect current workflow of Debian itself:
separates "stable" as distribution and as release, and adds testing
release. CDDs pays full respect to Debian work by now distorting it:
* What Debian consider testing is not promoted as stable by a CDD
* What Debian consider unreleased (except through testing snapshots)
is not promoted as released by a CDD
- Jonas
- --
* Jonas Smedegaard - idealist og Internet-arkitekt
* Tlf.: +45 40843136 Website: http://dr.jones.dk/
- Enden er nær: http://www.shibumi.org/eoti.htm
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
iD8DBQFH/fIzn7DbMsAkQLgRAvoCAJ440VW2pfWzAMw0e2CG2ex9ptlBsACfXcOS
26J9Ym8di7H3XhXBXQguRQ8=
=kA5W
-----END PGP SIGNATURE-----
Reply to: