1 |
On sob, 2017-05-06 at 19:36 +0200, Andreas K. Huettel wrote: |
2 |
> Am Samstag, 6. Mai 2017, 18:35:08 CEST schrieb Michał Górny: |
3 |
> > |
4 |
> > > An implementation is still missing though. |
5 |
> > |
6 |
> > ...and a rationale section to describe why you did the things this way. |
7 |
> |
8 |
> Well, that's more or less the Motivation section. Should I rename it? |
9 |
|
10 |
No. Motivation answers the question 'what is the problem being solved?', |
11 |
and in your case it serves that purpose well. Rationale is 'why did I |
12 |
choose this specific solution?' -- e.g. choice of file format, keywords |
13 |
and basically answers to every useful question that has been raised. |
14 |
|
15 |
> > 2. I can't say I like using magical keywords like 'testing' |
16 |
> > and 'unstable'; they're going to be confusing long-term (compare: |
17 |
> > the mess with stable/exp/dev for profiles). But I don't have a very good |
18 |
> > idea how to it better right now. |
19 |
> |
20 |
> Well, I pulled the two terms that are tradidionally used for ~arch... |
21 |
> "testing" and "unstable". Testing implied to me that a transition is taking |
22 |
> place, so that went to the "mixed state". |
23 |
|
24 |
I should point out that those terms are frequently used interchangeably, |
25 |
and adding disjoint meanings to them is least misleading. Perhaps a name |
26 |
like 'transitional' for the middle state would be better? |
27 |
|
28 |
> There's also Kent's proposal where some more indirection is introduced (see |
29 |
> the discussion threads). It more or less achives the same with even more |
30 |
> flexibility. I dont like it so much because I want to keep things simple. |
31 |
|
32 |
Yes, we don't need yet another python.eclass. |
33 |
|
34 |
> > 3. What is the use case for 'broken'? Are we ever going to use that? |
35 |
> |
36 |
> None that I know of. I only added it because it was suggested on the list (and |
37 |
> because it's simple to define and implement). |
38 |
|
39 |
Well, I was mostly asking because: |
40 |
|
41 |
a. without that, there would be no reason to repeat that dev/exp repoman |
42 |
magic in all three definitions, |
43 |
|
44 |
b. I fail to see any reason to have stable/exp/dev profile split for |
45 |
arch where even the most basic package integrity is not guaranteed. It's |
46 |
like having repoman check the profiles for nothing... |
47 |
|
48 |
-- |
49 |
Best regards, |
50 |
Michał Górny |