On Tue, Feb 03, 2004 at 05:30:19PM +0200 or thereabouts, Dan Armak wrote:
> I don't understand this comment. The developers would still work against a cvs 
> tree that contains all the latest stable stuff (base + changes) so why would 
> there be a problem with deps that wasn't in the orig GLEP?

Sorry -- I should have spelled this out a little more.

One of Spider's points was that people may not want to update every quarter
-- they might prefer an annual update cycle.  We're facilitating this by
guaranteeing ebuilds will be in the tree at least a year.  However, if
we're supporting distributions of the stable tree via tbz2s and
security/bugfixes via rsync, then I can see a problem with some of the
security/bugfixes requiring dependencies that aren't in some of the older
trees.

As a (purely hypothetical) example:

The 2004.0 stable tree gets released and includes OpenSSL 0.9.6

...11 months passes by...

A security vulnerability is found in gaim.  The new gaim ebuild is added to
the tree, but it depends on OpenSSL 0.9.7.  Now anyone using 2004.0 is
going to have problems.  We could work around this by including OpenSSL
0.9.7 in the security/update tree, but that also has a couple of problems:

* how can we easily figure out that OpenSSL 0.9.7 needs to get in there in
  the first place.
* It, in and of itself, isn't a security/bug fix update, yet anyone running
  the stable tree is going to get it as such the next time they sync their
  tree.  

--kurt

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to