1 |
Lie Ryan posted on Thu, 24 Dec 2009 08:08:48 +1100 as excerpted: |
2 |
|
3 |
> IMO Gentoo's edge was not about having the most cutting edge software |
4 |
> (pun not intended), but rather "having a choice". With Gentoo, you get |
5 |
> to choose which USE-flag to (not) include; you got to choose the kernel |
6 |
> options and also to use genkernel; then you've got a choice to run a |
7 |
> antiquated, full-stable, half-stable, ~arch, or overlay; you are free to |
8 |
> choose how antiquated or cutting edge you want your system to be. And |
9 |
> Gentoo's portage makes living the picky eater's life much easier than if |
10 |
> you have to compile packages and its dependencies manually to separate |
11 |
> the vegetables (or meats if you're a vegetarian; or pork if you're a |
12 |
> Muslim; or cows if you're a Hindi; or whatever taboo or personal |
13 |
> distrust you have). |
14 |
|
15 |
You're right about the choice, of course, but... well, the whole kde3 |
16 |
thing has nicely illustrated the issues stable gentooers have. |
17 |
|
18 |
To this day I'd not call kde4 ready for stable yet, and CERTAINLY not as |
19 |
stable and usable as kde-3.5.10. 4.4 should be getting close, I expect |
20 |
it'll be like a release candidate traditionally is, it could be stable if |
21 |
it had to be, but there's a few more bugs they want to kill before it's |
22 |
fully released. 4.3 is late beta, 4.2 was early beta, a LOT of SERIOUS |
23 |
bugs still hanging around, 4.1 was post-freeze alpha, and 4.0... was very |
24 |
early technology demo, mostly prototype, from a user perspective. |
25 |
|
26 |
OTOH, with the new name and focus on devs, KDE SC /is/ really aimed at |
27 |
devs, NOT end users, with the included apps really being developer demos, |
28 |
and the kde4 versioning and kde 4.2 stability claims /does/ more |
29 |
accurately reflect that -- it's just too bad they did the versioning so |
30 |
long before they announced their target audience change, as a lot of |
31 |
users were deceived into thinking it was ready for them... |
32 |
|
33 |
But be the upstream issues what they may, the problem for most |
34 |
distribution users including Gentoo users (and devs, BTW) is that support |
35 |
for the stable and production-ready version, kde3, ran out WAAYYYY before |
36 |
the next version, kde4, was similarly stable and production ready. |
37 |
|
38 |
"Oh, but there's the kde-sunset overlay." |
39 |
|
40 |
Yes, but it's officially user-only supported, that is gentoo-dev |
41 |
unsupported, because kde3 is unsupported upstream, as is the qt3 it's |
42 |
built upon, and there's no gentoo-devs interested in taking on the |
43 |
responsibility of continued support under those circumstances. That's |
44 |
not the sort of support stable users tend to be looking for. |
45 |
|
46 |
Meanwhile, the LTS/enterprise releases still have another year or more of |
47 |
kde3 coverage, as that's what was stable and shipping when their LTS |
48 |
product shipped (bar Ubuntu, of course, since they didn't ship an LTS |
49 |
kubuntu precisely because they foresaw exactly this sort of issues coming |
50 |
up, despite all the claims of continued support from kde at the time, |
51 |
claims that turned out to be worthless, for the ordinary distribution |
52 |
user -- but in hind sight kde was even then already refocusing their |
53 |
targetting, and weren't talking about the ordinary user any more). |
54 |
|
55 |
But back on the topic of Gentoo. Gentoo is and always has been a rolling |
56 |
upgrade community distribution, that reasonably closely follows |
57 |
upstream. When upstream drops support, Gentoo, without the resources of |
58 |
the enterprise/corporate distributions, has little choice but to |
59 |
ultimately drop support as well. Sure, the packages stay in-tree for |
60 |
awhile sometimes, but they don't actually build with modern gcc against |
61 |
modern system libs, and eventually, treecleaners or someone notices, and |
62 |
they get pulled. |
63 |
|
64 |
That's not the sort of thing stable users enjoy, for sure. Really, |
65 |
neither do they tend to enjoy the constant updates Gentoo has, changing |
66 |
their work environment out from under them. Good Gentooers soon learn |
67 |
that if they're updating less than once a month, the updates DO pile up, |
68 |
and the process DOES get rough. By three months, an upgrade gets |
69 |
difficult and stressfull, by six months, it's getting easier to start |
70 |
from a brand new stage-3, by a year, which is what Gentoo /does/ /try/ to |
71 |
support, a brand new stage-3 is generally going to be much easier than |
72 |
the exotic bugs you'll get trying to update in place. Yet stable users |
73 |
normally /want/ their stuff stable for a year or more, and expect no |
74 |
serious problems on update within their release slot, even a year or more |
75 |
out. The all-at-one-time release upgrade, OTOH, is assumed to be the |
76 |
normal case. Meanwhile, gentoo support for stale packages disappears |
77 |
rather soon, relatively, and users are forced into either not updating |
78 |
any more (no security updates) or upgrading. The enterprise/LTS |
79 |
distribution releases at least have a support timeclock that people can |
80 |
schedule their computing life around. |
81 |
|
82 |
As I mentioned above, it took the kde3/4 fiasco to really open my eyes to |
83 |
this, but open them it most certainly did! Generally speaking, |
84 |
enterprise and debian stable are the only ones supporting kde3 still, |
85 |
even tho kde4 isn't yet ready to fill its shoes for production machines. |
86 |
|
87 |
> For me, I run a mostly stable system and unmasks a few packages that I |
88 |
> used most frequently since those are the software that I have the time |
89 |
> to test thoroughly since I work with them all the time. I've been |
90 |
> running a python 3 overlay (very unstable at that time), but I'm not |
91 |
> willing to run a full ~arch since most of those software I don't use |
92 |
> often enough anyway. |
93 |
|
94 |
Of course, that's where Gentoo excels. It gives you the choice and |
95 |
ability to do just that, even if it's not that well supported. But in |
96 |
fact, because it's so easy and so necessary for stable users at times, |
97 |
there's /enough/ people doing it, that it generally works out |
98 |
/reasonably/ well. But still, tho the problems will be a bit different, |
99 |
I don't think running all ~arch is much different in overall problems |
100 |
than partial, or indeed, all stable, because if nothing else, hardware |
101 |
updates tend to bite all-stable people harder than all ~arch people, and |
102 |
also because stable /is/ a bit stale at times, and it's simply hard to |
103 |
remember what the fix was for that problem that happened over a year ago. |
104 |
|
105 |
I know I've certainly experienced that myself, running the kernel rcs, |
106 |
when the release is what goes ~arch, and stable is generally a release |
107 |
behind that. So when folks ask about kernel problems on the brand new |
108 |
stable kernel they're just upgrading to now, it's typically six months or |
109 |
more since I encountered the same issue, and I've often long since |
110 |
forgotten the details, as I'm on to newer and different problems. The |
111 |
full release would seem to be about right, I'd think, for most users not |
112 |
wishing to push the edge, as it's at least new enough the edge pushers |
113 |
still remember the issues and how to fix them, while being old enough the |
114 |
big issues all generally have fairly well known solutions. If I'm not |
115 |
mistaken (I run direct linus kernels and don't touch gentoo's kernel |
116 |
distribution at all, tho I know when they go stable since I follow the |
117 |
dev list and see the announcements/warnings there), current release is |
118 |
what gets ~arched for at least ~x86 and ~amd64 on Gentoo, so that's what |
119 |
I'd think would be about the best place to be, on a package I happen to |
120 |
follow reasonably closely, upstream. Similarly for a couple others I |
121 |
follow reasonably closely upstream. |
122 |
|
123 |
-- |
124 |
Duncan - List replies preferred. No HTML msgs. |
125 |
"Every nonfree program has a lord, a master -- |
126 |
and if you use the program, he is your master." Richard Stallman |