recent discussion regarding 'checksums'

[Benjamin R. Haskell]

On Tue, 28 Sep 2010, Matt McCutchen wrote:

> On Mon, 2010-09-27 at 22:33 -0400, Benjamin R. Haskell wrote:
>> But the flip side is that rsync is not a security tool.  MD5 is fine 
>> for rsync for the same reason SHA-1 (which, as with all hashes, will 
>> eventually be "broken") is fine for git:
>
> This gets a little off topic, but I /do/ want git to use a 
> collision-resistant hash function.  I would like to be able to fetch 
> from others without giving them a free pass to poison my repository. 
> I believe that was the original intended semantic of the "fetch" 
> operation; it's only now eroding as SHA-1 gets studied.  But the risk 
> isn't great enough to goad me into action yet.

Shoot.  I left out the link that was the whole point: it's Linus 
explaining why SHA-1 is fine (because security happens at a higher 
level):

http://kerneltrap.org/mailarchive/git/2006/8/28/211065

(Boils down to the two points I paraphrased -- but the upshot is that 
hash collision != free pass to poison)

-- 
Best,
Ben