| 1 |
On Thursday 25 March 2004 20:55, Chris Bainbridge wrote: |
| 2 |
> On Thursday 25 March 2004 19:22, Jon Portnoy wrote: |
| 3 |
> > The difference is that we (the developers) control our machines. |
| 4 |
> |
| 5 |
> Given that its possible to become a developer without any certification |
| 6 |
> process other than being able to fix a few bugs and use irc; who is really |
| 7 |
> in control? |
| 8 |
> |
| 9 |
> * Become a dev |
| 10 |
> * Upload trojan ebuild to randomly corrupt hd then rm -rf / after 24 hours |
| 11 |
> * Cackle as tens of thousands of systems are destroyed |
| 12 |
|
| 13 |
In the end it is allways another dev that asks someone to become a dev. In |
| 14 |
general people that offer themselves without a long previous track record are |
| 15 |
not taken all that serious. One does not become developer after fixing one or |
| 16 |
two bugs. This does not change the fact that indeed there are security issues |
| 17 |
involved. |
| 18 |
|
| 19 |
However those issues can only be solved by changing procedures within gentoo |
| 20 |
and are only sideways related to improving rsync security. In short, it is on |
| 21 |
our TODO list however it is not on top, while rsync security is. |
| 22 |
|
| 23 |
> Is it really that simple? And to fix it is so easy.. just keep a list of |
| 24 |
> people allowed to modify each directory. Developers sign, users check. |
| 25 |
|
| 26 |
That list is all devs with cvs commit access. As it is globally equal for now |
| 27 |
we only get one list instead of per-directory lists. Also your ACL proposal |
| 28 |
adds a lot of other features that may or may not enhance security. Security |
| 29 |
is a very difficult thing and a single mistake can easilly remove the |
| 30 |
security of the whole system. I remember that for example such a flaw existed |
| 31 |
in gpg, and another one in ssh. |
| 32 |
|
| 33 |
> I can't really understand this thread of conversation.. |
| 34 |
> |
| 35 |
> "Hey, heres a way of solving some security problems" |
| 36 |
> "We're not interested in solving all of those problems at the moment, just |
| 37 |
> one of them" |
| 38 |
We are interested in solving things fast. We are indeed not interested in |
| 39 |
solving all possible compromises, we are interested in an incremental path |
| 40 |
(small changes) where especially in the start we want to fight one issue at |
| 41 |
the time with as little organizational changes as possible. |
| 42 |
|
| 43 |
> "But you can fix the whole system, and its not difficult" |
| 44 |
Designing a system is not extremely difficult. Designing the system in such a |
| 45 |
way that it fits with the current gentoo procedures is impossible. Designing |
| 46 |
the system in such a way that the change to the procedures is mimimal is |
| 47 |
hard. Convincing all developers that those changes are needed and that they |
| 48 |
must accept that their privileges are taken away is very harder. Actually |
| 49 |
making all developers follow the new procedures is even harder. |
| 50 |
|
| 51 |
In all this keep in mind that the developers (and myself) are volunteers who |
| 52 |
will run away if they do not like how things are going. Currently not many |
| 53 |
developers have left, and we are not afraid to let developers go, but that |
| 54 |
does not mean that we can just ignore the developers. |
| 55 |
|
| 56 |
> "Not interested. We only want to fix one problem for now." |
| 57 |
|
| 58 |
We are interested, but would like problem 2 to 1000 to each be discussed in a |
| 59 |
separate thread as we want to fix the problem before april is over. If the |
| 60 |
discussion lasts untill half of April we are never going to be able to do |
| 61 |
this. There are still many issues on the implementation that need to be |
| 62 |
solved (and discussed) after the initial keyring is solved (and found secure |
| 63 |
and workable). |
| 64 |
|
| 65 |
Paul |
| 66 |
|
| 67 |
-- |
| 68 |
Paul de Vrieze |
| 69 |
Gentoo Developer |
| 70 |
Mail: pauldv@g.o |
| 71 |
Homepage: http://www.devrieze.net |
Archives