PHK's MD5 might not be slow enough anymore

[Chuck Swiger]

Hi--

On Jan 28, 2010, at 12:10 PM, Bill Moran wrote:
> This would also introduce a complete incompatibility between systems.
> I, for one, frequently copy password files from one system to another.
> I expect $1$ to be compatible on all systems.

Exactly.  Just like classic DES passwords were portable to all platforms.

> If a new algorithm is to be used, why even start with md5?  Why not
> start with something that's inherently stronger and more CPU intensive?
>> 
> From there, assign it a new algorithm number.  See the "Modular Crypt"
> section of crypt(3).  Then compatibility is maintained.

+1.  We're probably fine with MD5 password hashes against all but extreme measures for some time to come, but adding SHA-1 and being ready for whatever algorithm(s) might be chosen by NIST for SHA-3 would be a fine thing to do.

Regards,
-- 
-Chuck