| 1 |
On Sat, 19 Nov 2005 15:29:30 +0900 |
| 2 |
Jason Stubbs <jstubbs@g.o> wrote: |
| 3 |
|
| 4 |
> On Saturday 19 November 2005 15:01, Robin H. Johnson wrote: |
| 5 |
> > After my post to -core about how to move ahead with signing, I |
| 6 |
> > thought the next best place to continue is in a discussion of how |
| 7 |
> > Portage handles manifests and their signatures. |
| 8 |
> > |
| 9 |
> > First, the blatantly obvious, for the benefit of same developers, |
| 10 |
> > even though it's not relevant to signing. It is still a weak-point |
| 11 |
> > and does need to be addressed. Multiple-hashes! |
| 12 |
> |
| 13 |
> Yep, portages that don't break on multiple hashes being specified |
| 14 |
> have been around long enough for this to now be feasible. |
| 15 |
|
| 16 |
Hmm, sneak it in .53? *g* |
| 17 |
Just joking. |
| 18 |
|
| 19 |
> > So now the new Manifest structure looks roughly like this |
| 20 |
> > (abbreviated): -- PGP |
| 21 |
> > MD5 ... |
| 22 |
> > MD5 ... |
| 23 |
> > -- SIG |
| 24 |
> > -- SIG |
| 25 |
> > -- PGP |
| 26 |
> > MD5 ... |
| 27 |
> > -- SIG |
| 28 |
> > -- SIG |
| 29 |
> > etc. |
| 30 |
> > |
| 31 |
> > This has one important implication for backwards compatibility in |
| 32 |
> > checking of Manifests. |
| 33 |
> > In the case that a filename appears more than once in the file, only |
| 34 |
> > the last instance of it should be used, as that is the one that |
| 35 |
> > relates to the current version of the file. It's 4 lines of code in |
| 36 |
> > the current portage that need to be removed for this to work (see |
| 37 |
> > my -core post for where exactly). |
| 38 |
> |
| 39 |
> Hence, if we fix it in the next version we still have to wait six |
| 40 |
> months to a year for most everybody to be using it so we don't break |
| 41 |
> lots and lots of systems... |
| 42 |
> |
| 43 |
> Wouldn't it be easier to just disallow unsigned commits on the server |
| 44 |
> side? |
| 45 |
|
| 46 |
Probably, but might have an impact on server load (but Robin is the |
| 47 |
expert there ;). But that only covers one part of Robins problem |
| 48 |
(unsigned commits) but not the other (resigning of unchanged files), so |
| 49 |
he still needs the transactional Manifests, but I don't think the |
| 50 |
problem is worth the complications. |
| 51 |
|
| 52 |
Marius |
| 53 |
|
| 54 |
-- |
| 55 |
Public Key at http://www.genone.de/info/gpg-key.pub |
| 56 |
|
| 57 |
In the beginning, there was nothing. And God said, 'Let there be |
| 58 |
Light.' And there was still nothing, but you could see a bit better. |
Archives