http://lwn.net/Articles/307281/

the authors of Git seem to be willing to take the chance,
and willing to bet the Linux kernel on it.

If a Leo file was ever corrupted by a sha1 collision, it might
make for good press, the kind of thing slashdot loves.



On Fri, Jun 12, 2009 at 8:07 AM, Edward K. Ream<[email protected]> wrote:
>
> In the present hashcashe scheme, Leo will lose data if two external
> files have the same sha-1 hash: the file hashed later will *replace*
> the file hashed earlier.  The question is, how likely is such a
> collision?
>
> There are two conflicting points of view:
>
> 1.  The number of all possible hashes is 16**40 == 8 ** 80.  This is a
> truly enormous number, much bigger than the total documents that will
> ever be written in all of human history, no matter how long that turns
> out to be :-)
>
> 2. The number of all possible documents is (almost) infinitely
> larger.  For example, the number of all ascii files containing 1000
> characters is approx 128 ** 1000.  Thus, there would be *lots* of
> collisions **if** all such files were hashed.  There would be even
> more potential collisions in the set of all megabyte files.  And so
> on.
>
> Most discussions of sha-1 collisions focus on cryptanalysis attacks,
> and do not seem to be relevant.  Can anyone resolve the conflicting
> points of view?  It's important now for Leo, and may get even more
> important later.
>
> Offhand, I can think of now way to "recover" from an unexpected
> collision.  I suspect, but do know know for sure, that collisions
> would create havoc in a git repository.  Can anyone say anything for
> sure on this topic?
>
> Edward
> >
>

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"leo-editor" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/leo-editor?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to