1 |
Chris Gianelloni wrote: |
2 |
> On Thu, 2007-07-05 at 18:47 -0400, Mike Frysinger wrote: |
3 |
>> you proposing we rearchitect it all or just for testing purposes before going |
4 |
>> live ? i can see both ... |
5 |
> |
6 |
> I am proposing rethinking all of it. My current thoughts run something |
7 |
> like this: |
8 |
> |
9 |
> arch/amd64 |
10 |
> arch/ppc (not ppc/ppc64 or ppc/ppc32) |
11 |
> base |
12 |
> default/linux |
13 |
> default/freebsd |
14 |
> default/macos |
15 |
> kernel/darwin |
16 |
> kernel/linux |
17 |
> kernel/freebsd |
18 |
> release/2007.1 |
19 |
> target/desktop |
20 |
> target/server |
21 |
> userland (these aren't all the same type of thing) |
22 |
> userland/32-bit |
23 |
> userland/64-bit |
24 |
> userland/multilib |
25 |
> userland/freebsd |
26 |
> userland/hardened |
27 |
> userland/linux (this could be glibc, instead) |
28 |
> userland/macos |
29 |
> userland/no-nptl (not sure we really need this, at all) |
30 |
> userland/nptl (this either) |
31 |
> userland/selinux |
32 |
> userland/uclibc |
33 |
> |
34 |
> Of course, this is just my rough outline. What you would end up with, |
35 |
> as a profile, is something like this: |
36 |
> |
37 |
> default/linux/amd64/2007.1/desktop (not much different from now) |
38 |
|
39 |
I kinda thought up a system like this long ago, but it was more in line with |
40 |
node-based profiles. And wou;d've required gutting the current profile code in |
41 |
portage entirely. The idea being that, you construct the profile up in nodes |
42 |
from the top level (much like one does their PATH variable), and the profiles |
43 |
would be re-arranged into things like arch/, libc/, kernel/, etc.. In a way, I |
44 |
re-organized mips' 2007.1-dev profiles to quasi reflect how we'd look in such a |
45 |
layout. |
46 |
|
47 |
But I like this idea -- it goes halfway towards nodes to some extent (at least |
48 |
lines things up for nodes or some other implementation that maybe treats parents |
49 |
better). |
50 |
|
51 |
antarus even had a small draft document up on it that's better in detail: |
52 |
http://dev.gentoo.org/~antarus/essays/mixin-profiles.txt |
53 |
(later on, it was decided that there would have to be a pre-defined order for |
54 |
the first four nodes: base:arch:kernel:userland, and these first four nodes |
55 |
could not repeat. Everything thereon after was swappable and allowed to be |
56 |
placed in any order, such as base:mips:linux:glibc:ip30:o32 (where o32/ip30 can |
57 |
be swapped around)) |
58 |
|
59 |
But I definitely see this as a 2008.0 thing at the earliest. I also see no |
60 |
problem with mips joining in on the fun to play with things either! |
61 |
|
62 |
|
63 |
--Kumba |
64 |
|
65 |
-- |
66 |
Gentoo/MIPS Team Lead |
67 |
|
68 |
"Such is oft the course of deeds that move the wheels of the world: small hands |
69 |
do them because they must, while the eyes of the great are elsewhere." --Elrond |
70 |
-- |
71 |
gentoo-dev@g.o mailing list |