1 |
Rich Freeman: |
2 |
> On Wed, Jan 21, 2015 at 11:36 AM, Julian <hasufell@××××××××.de> wrote: |
3 |
>> |
4 |
>> We would need much better tools and PM support for overlays. I've talked |
5 |
>> about this too and there are examples of at least two distros that are |
6 |
>> running a more decentralized model: |
7 |
>> 1. NixOS, also see my thread on nix-dev about how they want to ensure |
8 |
>> focus [0] |
9 |
> |
10 |
> I still don't get how NixOS avoids some fundamental issues. As I |
11 |
> understand it they assign a unique ID to every build of everything, |
12 |
> and the dependencies are expressed in the same way. So, my build 123 |
13 |
> of appfoo v1.1 depends on build 456 of libz v2. Your package might |
14 |
> depend on build 457 of libz v2, and that is fine since both versions |
15 |
> will be installed side-by-side. Then we get to have two copies of |
16 |
> libz in RAM and maybe build 456 has a security vulnerability. It is |
17 |
> better than static linking, but not by far. |
18 |
> |
19 |
|
20 |
All copies are known to the PM. That's nothing like bundles libs or |
21 |
static linkage. It's probably a triviality to develop a tool that checks |
22 |
for vulnerable libraries (not sure if there already is one). |
23 |
|
24 |
I didn't say we should become NixOS. It was just one example how |
25 |
decentralized packaging CAN work. There are other examples. |
26 |
|
27 |
> |
28 |
>>> How |
29 |
>>> can we have both games.eclass and no games.eclass but due to a |
30 |
>>> technical/methodology change there are now less problems? Ditto for |
31 |
>>> the two multilib implementations? |
32 |
>>> |
33 |
>> |
34 |
>> Well, for once: forking an overlay is easier than forking the whole |
35 |
>> gentoo tree. |
36 |
>> But that's not necessarily the main point. Ideas would not be decided by |
37 |
>> "who does something first in the tree", but by more dynamic processes |
38 |
>> about approval in the community. Someone starts a repo and does stuff. |
39 |
>> If people like it, they are going to use it and focus their contribution |
40 |
>> there. |
41 |
> |
42 |
> So, which is it? You point out that we have inconsistencies in our |
43 |
> tree because we have games.eclass but we don't force everybody to use |
44 |
> it. |
45 |
> |
46 |
> How is it better if the games are scattered across 47 overlays |
47 |
> instead, and some of them use games.eclass, others use something else |
48 |
> that sticks games in an entirely different group/path/whatever, others |
49 |
> follow upstream, and maybe half the overlays do security updates and |
50 |
> the other half don't? Maybe half of them use the multilib eclass and |
51 |
> half use portage-multilib. |
52 |
> |
53 |
> So, are we OK with inconsistencies or not? If we are OK with them, |
54 |
> then why complain about it? |
55 |
> |
56 |
|
57 |
It is better, because you can actually choose what you want without |
58 |
messing with complex masks and workarounds. |
59 |
And because people could even set up monolithic repositories (actual |
60 |
distros) that merge the work of several overlays (that's what happens in |
61 |
NixOS afaiu, some companies seem to be interested in that work too), |
62 |
straightening out things, checking for compatibility, removing |
63 |
overlapping parts, etc. |
64 |
People already use overlays so extensively that I don't think the world |
65 |
would collapse from the user perspective. And keep in mind that a lot of |
66 |
gentoo projects already moved their main work to github, because they |
67 |
cannot keep up the massive amount of work without community contributors. |
68 |
|
69 |
I think modularity does not only make sense for programming in general, |
70 |
but also for packaging. But there must be a concept behind it as well, |
71 |
otherwise you just get terrible fragmentation and ::mynick overlays (see |
72 |
exherbo). |
73 |
|
74 |
The main point about decentralized packaging is to shrink the tree (and |
75 |
needed manpower) and allow more community effort to shape the "distro" |
76 |
while still maintaining QA through a review workflow (to not become the |
77 |
next arch linux user repository, aka AUR). |
78 |
Then you can focus on what is most important: PM, EAPI, toolchain maybe |
79 |
and most importantly... review. |
80 |
|
81 |
But I've said most of this already. |
82 |
|
83 |
>> |
84 |
>>> I think that a major shakeup only makes sense if we can actually |
85 |
>>> demonstrate a new model in operation, and it actually solves our |
86 |
>>> problems. |
87 |
>>> |
88 |
>> |
89 |
>> I've already given two examples of a new model in operation. Also, I |
90 |
>> haven't said "let's do this tomorrow". IMO it will never happen anyway. |
91 |
>> I am convinced that gentoo will rather die slowly, because people are |
92 |
>> afraid to change the course which leads us to a nice thick wall, because |
93 |
>> we might derail. |
94 |
>> I'm still not sure if it will die because of "lack of manpower" or |
95 |
>> because our technology is so messed up that it becomes unusable. |
96 |
> |
97 |
> So, I'm not convinced that we won't find our way, but if Gentoo stops |
98 |
> existing because we've all found a better way and moved on to it, then |
99 |
> really there is no loss unless you're really attached to the name, |
100 |
> "Gentoo." None of us own stock in the Gentoo Foundation. If |
101 |
> something new already exists and is doing it better, and we all just |
102 |
> decide to switch, then we all reap the benefits. |
103 |
> |
104 |
> However, it seems likely to me that if there were an overwhelming |
105 |
> sense that a new model would really work for most of us, we'd change. |
106 |
> The problem is that not everybody is here for the same reason, and |
107 |
> we're best off trying to tap into that rather than fighting it. |
108 |
> |
109 |
|
110 |
I don't have a completely worked out idea for gentoo. I couldn't even |
111 |
write a GLEP about this and I certainly won't, because it would be a |
112 |
huge waste of time. The level of enthusiasm is just too low and any such |
113 |
thing cannot and will not happen on GLEP level. It would have to be a |
114 |
process, a community effort, not something the council decides on and |
115 |
says "ok, now we can implement it" and would imply studying other |
116 |
distros, checking out their workflow etc. |
117 |
|
118 |
And btw., this was only one of two proposals. |
119 |
|
120 |
My point was: it's not enough talking about "what should we focus on?". |
121 |
How do you implement that focus? |
122 |
|
123 |
What's your idea there, except of keeping status quo which clearly does |
124 |
not work since even gentoo oldtimers seem to realize something is going |
125 |
wrong? |