1 |
On Sat, 14 Jun 2008 12:45:31 +0200 |
2 |
Luca Barbato <lu_zero@g.o> wrote: |
3 |
> > And none of those are even close to a reasonable, implementable |
4 |
> > idea. |
5 |
> |
6 |
> They are implementable. |
7 |
|
8 |
They're really not. You haven't even begun to discuss: |
9 |
|
10 |
* What generation looks like. |
11 |
* How to select which ebuilds to trigger generation for. |
12 |
* When specifically to trigger generation. |
13 |
* Whether generation failure is possible, and what happens if it is. |
14 |
* What to do when generated information is required but not available. |
15 |
* The security implications of the previous point. |
16 |
* What the impact upon upstream servers is, if any. |
17 |
* How generation fits in with users doing system updates. |
18 |
* How generation fits in with the highly differing frequencies of |
19 |
upstream updates. |
20 |
* How generation can be made to fit in without changes to the version |
21 |
spec, whilst still 'making sense' and doing the right thing. |
22 |
|
23 |
The answers to all of the above are highly non-trivial. |
24 |
|
25 |
-- |
26 |
Ciaran McCreesh |