1 |
Duncan wrote: |
2 |
|
3 |
> Steve Long <slong@××××××××××××××××××.uk> posted |
4 |
> ekol7b$q8i$1@×××××××××.org, excerpted below, on Fri, 01 Dec 2006 07:23:09 |
5 |
> +0000: |
6 |
> |
7 |
>> Excellent; pkgcore really sounds great- is there any possibility that |
8 |
>> it'll become the new portage? |
9 |
> |
10 |
> Possibility, yes. It's not certain, as there are multiple contenders |
11 |
> (paludis is the other), and it will be some time, in any case. |
12 |
> |
13 |
> The current problem is that there's no standard definition for what |
14 |
> constitutes an acceptable ebuild, beyond the basic gentoo dev guidelines. |
15 |
> The de facto definition is whatever works with versions of portage |
16 |
> currently in the tree (or just barely removed), but that presents many |
17 |
> difficulties, including both slow upgrades since backward compatibility |
18 |
> must be maintained for longer even when the former functionality is |
19 |
> considered b0rken, and questions of what's broken, the package manager or |
20 |
> the ebuild, when something fails to work as expected. |
21 |
> |
22 |
I'd vote for the defacto with strict backward compatibility, and perhaps a |
23 |
directive/ alias for newer scripts. If something really doesn't work and |
24 |
someone cares (bug reporter), ask them to update the ebuild if needed. So |
25 |
long as good docs are in place (the dev handbook I've seen somewhere is an |
26 |
example) for the update process, that's acceptable in my book. |
27 |
|
28 |
> Thus, all three package managers, the current portage solution, and |
29 |
> paludis and pkgcore as well, are currently under slower development than |
30 |
> they might otherwise be, while interested parties attempt to hash out a |
31 |
> working standard definition of what actually constitutes a proper ebuild, |
32 |
> and what helper functions said ebuild can in fact depend upon the package |
33 |
> manager to make available. Once that's decided and approved, the playing |
34 |
> field upon which the merits of the next generation package managers can be |
35 |
> judged will be much fairer for all. Of course, with that defined, portage |
36 |
> itself will be freer to progress at speed as well, and it may be that it |
37 |
> will remain the default "approved" solution for quite some time. |
38 |
> |
39 |
As for helper functions, I'd guess a union of all available ;) |
40 |
|
41 |
|
42 |
-- |
43 |
gentoo-dev@g.o mailing list |