Re: [Monotone-devel] Re: Hash collisions resiliency

[Nathaniel Smith]

On Thu, Apr 14, 2005 at 07:04:26AM -0400, Frank Ch. Eigler wrote:
> 
> Nathaniel Smith <address@hidden> writes:
> 
> > [...] Detecting a problem: this depends on what collides.  [...]
> 
> Perhaps concerned users might be somewhat reassured by an experiment
> such as the following.  Create a specially crippled version of
> monotone, which uses (say) only 12-bit hashes.  Run whatever test
> suite you can, a bunch of times.  See what blows up.  Add any new
> practical consistency checks that those booms suggest would help.

This is kinda a silly idea?  What will blow up is that at netsync
time, monotone will simply not realize that it is missing some data,
and silently replace the data it is supposed to fetch with data it
already has.  There's no practical consistency check against this.

(And besides, "seeing what blows up" is kind of the wrong way to go
about it, since it's the parts that blow up that you don't need to
worry about --- it's silent corruption that makes people nervous...)

-- Nathaniel

-- 
  /* Tell the world that we're going to be the grim
   * reaper of innocent orphaned children.
   */
-- Linux kernel 2.4.5, main.c