From: Nathan Loofbourrow <loofbour@cis.ohio-state.edu>
Date: Mon, 9 May 94 21:17:37 PDT
To: cypherpunks@toad.com
Subject: Re: PGP 2.5
In-Reply-To: <9405092209.AA21090@nyx10.cs.du.edu>
Message-ID: <199405100404.AAA06262@styracosaur.cis.ohio-state.edu>
MIME-Version: 1.0
Content-Type: text/plain


Paul Grange writes:
 > |> Another RSAREF limitation is that it cannot cope with keys longer than
 > |> 1024 bits.  PGP now prints a reasonably polite error message in such a
 > |> case.
 > 
[...]
 > This restrcition comes from RSAREF code, over which the PGP team had no 
 > control.

Strange -- the RSAREF 2.0 license asserts no such restriction, unless
I've misread it. Patching it -- say, to allow it to handle >1024 bit
keys -- would seem to fall under one's license...

[from license.txt]
     c.   to modify the Program in any manner for porting or
          performance improvement purposes (subject to Section 2)
          or to incorporate the Program into other computer programs
          for your own personal or internal use, provided that you
          provide RSA with a copy of any such modification or
          Application Program by electronic mail, and grant RSA a
          perpetual, royalty-free license to use and distribute such
          modifications and Application Programs on the terms set
          forth in this Agreement.

Is the definition of "performance improvement" so limited that
improving maximum key size is not permitted?

This aside, modifying RSAREF 2.0 (and taking out the guardrails in
keymgmt.c) *appears* to allow larger key sizes. The only succeeding
restriction on key sizes is the 1280-bit restriction imposed by the
assembly code, if the comments are to be believed.

Generating a brand new ~1280 bit key under 2.5 appears to work
perfectly, although I suppose RSAREF could be happily returning a
shorter key that claims to be >1024 bits (either by design, or by
omission). The fact that an older >1024 bit key fails this test does
raise this suspicion.

This will take some further work. I would be surprised to discover
that the MIT folk hadn't fiddled with this at all, though -- Any
comment from the 2.5 folks on the barriers to using RSAREF for longer
keys?

nathan