1 |
Rich Freeman wrote: |
2 |
> something is bound to break for good sooner or later if things don't change. |
3 |
|
4 |
Certainly. |
5 |
|
6 |
But consider the chain of events: |
7 |
|
8 |
* user is happily using outdated, but working, software without |
9 |
knowing how behind the times upstream really is, because it works |
10 |
* gentoo masks and removes package |
11 |
|
12 |
That looks bad. |
13 |
|
14 |
Masking is a quite invasive UX. It makes a package unavailable by |
15 |
default *before* the package actually stops working. |
16 |
|
17 |
Users who want to fix the problem need to get involved upstream, |
18 |
there is no disagreement about that, but users who have already |
19 |
gotten a package masked by the powers that be are vastly less |
20 |
motivated to do so, because the package has already been masked. |
21 |
|
22 |
Masking communicates that a decision to treeclean has been made. |
23 |
|
24 |
Masking is not at all communicating an opportunity to address the |
25 |
perceived problems. That should be done in a different way. |
26 |
|
27 |
A per-ebuild bug metric would be cool. A kind of health indicator |
28 |
for individual ebuilds, alerting users when some of our installed |
29 |
ebuilds go yellow, so that we have perhaps on the order of six |
30 |
months before the package goes red, at which point it would be fine |
31 |
to mask at will. Does that make sense? (Obviously how many months |
32 |
yellow is depends on what else happens in the tree. It's a ballpark |
33 |
figure.) |
34 |
|
35 |
|
36 |
//Peter |