This is the mail archive of the libc-hacker@cygnus.com mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: 2.2 projects


> cc: libc-hacker@cygnus.com
> Date: Fri, 29 Jan 1999 23:54:34 -0500
> From: Zack Weinberg <zack@rabi.columbia.edu>
> 
> On Sat, 30 Jan 1999 15:41:28 +1100, Geoff Keating wrote:
> >> Date: Thu, 28 Jan 1999 11:15:50 -0500
> >> From: Zack Weinberg <zack@rabi.columbia.edu>
> >> 
> >> - OpenBSD extended crypt() [need non-US hacker]
> >
> >This is the Blowfish '$2a$' crypt prefix, isn't it?
> >
> >I don't really want to put that in, because it isn't exportable and we
> >already have a MD5-based crypt that is secure and exportable.  I guess
> >you could add a SHA-based crypt if you wanted an alternative to
> >MD5---it might be helpful to US government people who need to use
> >an approved algorithm.
> 
> I don't actually know how the OpenBSD stuff works.  I was thinking
> that the DES-based crypt() has some limitations that can be obnoxious
> from a sysadmin's point of view - e.g. 8 char cleartext limit - and
> that it's an old clunky algorithm that can be brute-forced these
> days.  OpenBSD is usually the place to look for security innovations,
> so...

Yes.  That's why the manual says that the DES-based crypt should not
be used, and is only supplied for compatibility (I'm sure this is news
to some people, which is why I wrote the manual).

I think the MD5-based crypt comes from OpenBSD too.  They prefer the
Blowfish version now, because it is more configurable, you can specify
how long you want to spend computing the hash to make dictionary
searches harder.

> Another thing.  I've seen it asserted that crypt() is exportable since
> it can only be used for authentication - e.g. Sun ships the same libc
> in domestic and exported Solaris.  Could we get away with that
> argument?  Has anyone gotten a hard legal opinion?

This may well be true, but it won't help much because the crypt
library also includes a DES API for the RPC stuff and for
compatibility (the `encrypt' functions), which is externally visible.
There may also be a problem exporting source code.

I think that it is possible to get a licence to export 56-bit crypto
(eg. DES) now anyway.  If someone in the US is bored, they might try
applying and see what happens.

-- 
Geoffrey Keating <geoffk@ozemail.com.au>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]