1 |
>On Thu, 24 Apr 2003, Joshua Brindle wrote: |
2 |
> |
3 |
>> there is no easy way since the only way cvs knows to allow/disallow commits is |
4 |
>> by permissions, we use permissions but they aren't fine grained, ie: everyone |
5 |
>> who has access to commit any package can commit to all of them. This is a lot |
6 |
>> better anyway since we have to be able to add new packages, do quick bumps |
7 |
>> on packages we don't necessarilly maintain, etc. Obviously if a dev is abusing |
8 |
>> we'll have records of what was commited and where and be able to take care |
9 |
>> of that. |
10 |
> |
11 |
>I maintained a CVS repository at my old workplace. |
12 |
>By taking advantage of the CVS passwd file, I was able to both avoid authenticating with local |
13 |
>users, *and* allow arbitrarily fine-grained access to the repository. |
14 |
> |
15 |
>If you think it would be useful, I could sketch out our permission scheme. It wasn't terribly |
16 |
>complicated, although it did result in creation of a lot of groups and users... |
17 |
> |
18 |
|
19 |
AFAIK this necessitates use of cvs pserver. We don't use pserver, it's less secure than |
20 |
cvs over ssh which is what we use. |
21 |
|
22 |
and when i said it can't be done easily i meant logistically, devs move around a lot, |
23 |
take care of what needs taken care of, etc. It would be a full time job keeping |
24 |
permissions up to date, and it also wouldn't be a productive use of time. Moreover |
25 |
we do have a need to authenticate locally since cvs isn't the only resource developers |
26 |
use on the machine. |
27 |
|
28 |
-- |
29 |
gentoo-hardened@g.o mailing list |