Re: MD2, MD5, etc

James M Galvin <galvin@tis.com> Thu, 09 April 1992 01:51 UTC

Message-Id: <9204081731.AA10244@TIS.COM>
Reply-To: James M Galvin <galvin@tis.com>
To: Jordan Brown <jbrown@qdeck.com>
Cc: ietf@isi.edu
Subject: Re: MD2, MD5, etc
In-Reply-To: Jordan Brown's message of Tue, 07 Apr 92 12:15:18 PDT. <9204071215.aa10825@sco1.qdeck.com>
Date: Wed, 08 Apr 1992 13:31:28 -0400
From: James M Galvin <galvin@tis.com>

	Why do we need two (three?  five?) such algorithms?  I will
	admit to almost complete ignorance of the topic, but when the
	algorithms can be described by exactly the same paragraph, it
	seems like endorsing more than one algorithm is a road to
	confusion, excess work, and so on.

Well, speaking for myself, the publication of the documents as
informational RFCs is not an endorsement per se, it is a distribution of
the specification of each algorithm.

As to why there is more than one, no single algorithm is suitable for
all situations.  In particular, each has different performance and design
characteristics.  In order from slowest to fastest we have MD2, MD5, and
MD4.

Check out the security considerations sections of MD4 for a statement
about its usage and design characteristics.  MD5 was created to balance
the performance of MD4 with the traditional design characteristics of
MD2.

MD2 is required for PEM while MD5 is required for SNMP Security and PPP
Authentication.  MD4 was considered for SNMP Security and PEM, but its
security is still being reviewed.  Keep in mind it is on the leading
edge of design characteristics for such algorithms.

Jim

Re: MD2, MD5, etc James M Galvin
MD2, MD5, etc Burt Kaliski
Re: MD2, MD5, etc Dave Katz