1 |
On Sun, Dec 8, 2013 at 2:14 PM, Ulrich Mueller <ulm@g.o> wrote: |
2 |
> |
3 |
> PMS just provides a mechanism, but doesn't prefer one SLOT value over |
4 |
> another. Such a change would introduce policy into PMS which is not |
5 |
> the right way to go. |
6 |
|
7 |
Sure it does - it defaults to :* when :* was never specified. I don't |
8 |
see how defaulting to :0= is a "policy" any more than :* is. |
9 |
|
10 |
> |
11 |
> If a dependency on a specific SLOT value is needed then it should be |
12 |
> explicitly specified in the ebuild. |
13 |
|
14 |
Honestly, I think this is kind of like saying that garbage collection |
15 |
is useless because programmers should just correctly free anything |
16 |
they create exactly once. |
17 |
|
18 |
If maintainers were generally giving careful thought to slots in |
19 |
dependencies then we wouldn't have packages that stick the slot in the |
20 |
package name instead. Sure, we can just ban packages like these and |
21 |
force everybody to fix all the breakage that results (which in theory |
22 |
should never have existed), but it seems better to me to try to make |
23 |
the best default the default. |
24 |
|
25 |
I guess we could just ban any non-explicit slot version dependency (ie |
26 |
90% of our current dependency atoms are now invalid), but that doesn't |
27 |
really seems a bit like programming in Ada. :) |
28 |
|
29 |
Rich |