From: Yusuke ENDOH <mame@...>
Date: 2011-04-16T21:39:26+09:00
Subject: [ruby-core:35784] Re: hashes are not consistent across ruby processes?

Hello,

IIRC, the behavior was changed with security in mind because
a guessable hash value is a possible cause of Denial of Service
attack.  I think I've heard that such an attack had been done
actually in Perl, but I have no pointer, so I could be wrong.


2011/4/16 Roger Pack <rogerdpack2@gmail.com>:
> Hello all.
> I have a small program here that caches lookups. �To do so it writes
> out a "unique" file based on the current ruby script, ex:
>
> filename = File.expand_path("~/#{$0.hash}")
>
> This worked great for looking up the file in 1.8.x, however in 1.9.x I
> see the following behavior:
>
> c:\dev>irb
>>> 'abc'.hash
> => -753597542
>>> 'abc'.hash
> => -753597542 # so it is consistent within the same process
>
> now run irb again:
>
> c:\dev>irb
>>> 'abc'.hash
> => 552773397 # different now!
>
> The hash values change each time you run a new process. �I would have
> expected them to stay consistent so this is a bit surprising (even the
> hash values for things like an empty array are different now, too, and
> didn't used to be).
>
> Any feedback on this?
> Thanks.
> -r
>
>



-- 
Yusuke Endoh <mame@tsg.ne.jp>