1 |
On 8/6/21 2:37 PM, n952162 wrote: |
2 |
> On 8/6/21 2:16 PM, Rich Freeman wrote: |
3 |
>> On Tue, Aug 3, 2021 at 2:03 AM n952162 <n952162@×××.de> wrote: |
4 |
>>> Well, what you say is likely true, but does "old software" really need |
5 |
>>> to be kept working? Couldn't problems necessarily only be dealt with |
6 |
>>> in the newest versions? |
7 |
>>> |
8 |
>> I think you are misunderstanding what actually went wrong in your |
9 |
>> situation. Nothing broke in your existing software. |
10 |
>> |
11 |
>> You're using an old version of portage. It will continue to work as |
12 |
>> it always has. |
13 |
>> |
14 |
>> However, you wanted to use it with a newer version of the software |
15 |
>> repository. This contained a package that wasn't compatible with old |
16 |
>> versions of portage. The version of portage you're using detected |
17 |
>> this, and refused to install it, so as to not randomly break your |
18 |
>> system. Your system continued to work as it always had. You just |
19 |
>> couldn't install that particular package, or anything that depends on |
20 |
>> it. |
21 |
> |
22 |
> |
23 |
> I think that doesn't characterize my point quite. |
24 |
> |
25 |
> I was complaining, mostly, that isodate had to be the thing that was |
26 |
> incompatible with my configuration. Maybe there is a unavoidable reason |
27 |
> that that package had to move to the newest EAPI, or maybe it was just a |
28 |
> sense that it's cool to be with the cutting edge. It seems to me that |
29 |
> isodate (which is actually tied, perhaps indirectly, to clearly slow |
30 |
> United Nations rule-making) must be pretty stable. |
31 |
> |
32 |
> |
33 |
>> |
34 |
>> Generally we try to maintain a reasonably sane upgrade path going back |
35 |
>> maybe six months or so. You just needed to update portage first. |
36 |
> |
37 |
> |
38 |
> My update was two months late. |
39 |
> |
40 |
> |
41 |
>> |
42 |
>> If your system is more than a month or two out of date just running |
43 |
>> emerge -uD world or whatever blindly is more likely to run into a |
44 |
>> problem. It shouldn't break your system unless you go adding random |
45 |
>> options to the command line to override safety features, |
46 |
> |
47 |
> |
48 |
> I don't have the expertise to do something like that. |
49 |
> |
50 |
> |
51 |
> |
52 |
>> but it might |
53 |
>> involve a few steps (like updating portage, @system, and so on before |
54 |
>> trying to update everything). |
55 |
> |
56 |
>> |
57 |
>> It usually isn't unmanageable, but Gentoo is definitely not an |
58 |
>> LTS-oriented distro. If you want to only get security fixes for three |
59 |
>> years and then update everything in one go, then stick with something |
60 |
>> like RHEL, Ubuntu LTS, or debian stable. Those distros deliver |
61 |
>> exactly that sort of experience, and really there isn't as much |
62 |
>> benefit to something like Gentoo if you're just going to update it |
63 |
>> every other year anyway. |
64 |
> |
65 |
> |
66 |
> Those distros are not source distros. I'm making an argument that |
67 |
> there's a large space between binary distributions and source |
68 |
> distributions that pass every upstream change down in realtime. Gentoo |
69 |
> is in the best position to service that space |
70 |
> |
71 |
> |
72 |
> |
73 |
|
74 |
I see that that begs the question of why someone would be interested in |
75 |
a source distribution. |
76 |
|
77 |
With a binary distribution, you fly on trust. With a source |
78 |
distribution, you do, too, given the size of the code base. |
79 |
|
80 |
The difference is, with a source distribution, there's a "paper trail" |
81 |
that can be followed back if that trust was abused. |
82 |
|
83 |
That's just about impossible with a binary distro. |