| 1 |
On Friday 09 July 2004 1:15 pm, John Davis wrote: |
| 2 |
> But at this point in time, Gentoo is *not* an Enterprise system. |
| 3 |
> Packages have zero consistency between releases (even between weeks). |
| 4 |
> Creating a whole new profile and then going through the hoops |
| 5 |
> (documentation and user support) to get them all to switch over to the |
| 6 |
> new profile far outweighs the benefit .. especially for (basically) one |
| 7 |
> package! |
| 8 |
|
| 9 |
1/ Saying that Gentoo isn't an enterprise system isn't doing anything to |
| 10 |
solve the problem. The fact remains we should be doing new profiles on a |
| 11 |
regular basis, and finding out what we need to get that done. |
| 12 |
|
| 13 |
2/ Package inconsistency is more justification for new profiles. We have |
| 14 |
had discussions earlier about separate branches in CVS, pinned packages in |
| 15 |
profiles, etc. Also see GLEP19. This is not even something that is |
| 16 |
Enterprise-specific, that's good QA. We shouldn't even have profiles at |
| 17 |
all if we don't use them. |
| 18 |
|
| 19 |
3/ Users should be able to settle into a profile for while, and only get |
| 20 |
occasional updates associated with security fixes and major bugs. |
| 21 |
Besides, it makes less sense for us to have old profiles that are moving |
| 22 |
targets -- the support overhead is greater juggling multiple profiles that |
| 23 |
are always changing, than freezing the old ones and focusing primarily on |
| 24 |
adding/updating packages in the upcoming release/profile. They should be |
| 25 |
more or less frozen, so that we know a certain combination of packages |
| 26 |
*works* beyond any reasonable doubt. |
| 27 |
|
| 28 |
We can't be having bikeshed debates in -dev forever. We've only had a |
| 29 |
general consensus that, yes we need some kind of Enterprise Gentoo... |
| 30 |
overlooking the fact that everyday users also need some kind of |
| 31 |
predictable release process and profiling. |
| 32 |
|
| 33 |
Yes, we'd need more people, and to some extent, more red tape. This has |
| 34 |
been an on-going discussion since even before I became a dev (2002) -- and |
| 35 |
we're no closer to implementing anything. |
| 36 |
|
| 37 |
xorg is a significant change. Ideally we'd be rolling up a number of |
| 38 |
significant changes into a new profile, and freezing the old one out. |
| 39 |
We're not there yet, but it's no reason to not try. |
| 40 |
|
| 41 |
> |
| 42 |
> The problem of consistency is far beyond the xfree/ xorg switch. Why put |
| 43 |
> all of the effort into one package when we don't have more important |
| 44 |
> things, like our toolchain, stabilized between releases? |
| 45 |
|
| 46 |
You know, there are a group of people working on toolchain. If the |
| 47 |
toolchain herd needs help to make things stable, they should get more |
| 48 |
people. That needs to go to devrel and they need to scale. That issue |
| 49 |
is independent of QA/Release Management/profiles/etc. |
| 50 |
|
| 51 |
Cheers, |
| 52 |
Dylan Carlson [absinthe@g.o] |
| 53 |
Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x708E165F |
| 54 |
|
| 55 |
-- |
| 56 |
gentoo-dev@g.o mailing list |
Archives