| 1 |
On Wed, 23 Oct 2019 01:16:51 -0400 |
| 2 |
Joshua Kinard <kumba@g.o> wrote: |
| 3 |
|
| 4 |
> And for Perl or Python, I think we should be making an effort to leverage |
| 5 |
> their respective mirroring systems first before putting their distfiles onto |
| 6 |
> our mirrors. Perl's got CPAN, and Python has pypi. For things that don't |
| 7 |
> exist on those systems, then we use our mirrors. |
| 8 |
|
| 9 |
We still have to mirror them, because upstream has a tendency to nuke |
| 10 |
things so that they can't be fetched any more from these primary |
| 11 |
sources. |
| 12 |
|
| 13 |
So whether end user fetch from the distfiles mirror for the first hit, |
| 14 |
or as a fallback, the cost is still there. |
| 15 |
|
| 16 |
The packages aren't broken, upstream hasn't stopped shipping it, just |
| 17 |
some upstreams have a fetish for nuking everything but the |
| 18 |
latest-and-greatest, and at a pace that is absolutely rediculous and |
| 19 |
can't be imagined for us to keep up with with all the stabilization |
| 20 |
rigmarole. |
| 21 |
|
| 22 |
Yes, backpan does exist, but its neither perfect, nor fast. |
| 23 |
|
| 24 |
And the faster upstream nukes things, the more likely it is it won't |
| 25 |
even be mirrored on backpan! |
| 26 |
|
| 27 |
( I wish I was imagining this circumstance, but its happened far too |
| 28 |
often ) |
| 29 |
|
| 30 |
And we're not doing our users any service by burdening them with this |
| 31 |
madness. |
Archives