[lug] MD5 strength?

[D. Stimits]

rm at fabula.de wrote:
...
>   - is MD5 considered safe? (not really, it takes _much_ more hard-
>     ware to find words that map to the same digest when using SHA).

This is the main point I am asking about. But, despite SHA requiring 
much better hardware and being much better than MD5, SHA-1 does not seem 
to be part of glibc, whereas MD5 is. My current understanding of MD5 is 
what I want to verify, I don't care about anything that glibc does not 
provide (libcrypt is part of glibc package, at least on RH, and probably 
all systems with glibc and "strong" passwords). A more precise question 
is to ask: If a shadow password file is encrypted with MD5, and a buffer 
overflow or other attack allows this file to be emailed or viewed by the 
attacker, is it correct or incorrect to say that MD5 might as well not 
even be used if the attacker is well-informed and has good home 
computing power (e.g., a quad Xeon, and 100 GB of extra disk space)? If 
the passwords can be broken, then how long will it take on typical 
hardware (e.g., the best single cpu Athlon currently available, or best 
P3 or P4)? My belief is that single DES is a joke, only casual observers 
are fooled by it...but that MD5 is a significant barrier. [SHA-1 is a 
big barrier, but anything not in glibc is to be ignored here]

D. Stimits, stimits AT attbi.com