Martin Packer wrote:
>I don't know when restrictions on encryption were lifted but when I first
>was involved with encryption in the late 1980's it was pretty restrictive
>who could have it.
>So the point is - because of the restricted availability - it's possible
>the injection of encryption into sites and applications might be less than
>desirable.

U.S. export restrictions on encryption were never mainframe-specific, and
they involved export, not the U.S. market. When they existed, the
restrictions affected practically every system. Former President Clinton
lifted most of the export restrictions in 1996 (with Executive Order
13026). But (as many observers pointed out at the time, and exploited) the
U.S. government never could legally block original source code published in
an exported, paper book, including source code that could be scanned (or
re-keyed), compiled, and run on a MVS (or OS/390) system of that era. Nor
could the U.S. government legally block anything manufactured outside the
United States by a non-U.S. entity.

The "fancy" stuff, e.g. secure key cryptography, was more tightly
controlled and is still controlled to some extent. (And not only by the
U.S. government.) Also not unique to mainframes, although mainframes do the
fancy stuff exceptionally well when they're allowed. Most systems don't do
the fancy stuff at all, which is a great reason to buy a couple IBM
mainframes (with CryptoExpress and, preferably but optionally, the TKE
workstation).

Mainframe application changes are rarely required to implement effective
encryption, and they will be even less often required over time if
encryption is the only requirement. I can't really say the same thing about
other platforms, not to that extent.

Encryption always has *some* cost. It's computational math, after all, plus
a bit of work (labor) to implement and to maintain. Not much cost,
especially nowadays with competent implementation, but nonzero. Security
breaches also have costs.

If somebody wants to make a programming language-centered argument about
security vulnerabilities, you'd find quite a bit of agreement that C and C+
+ are comparatively weak foundations from a security point of view, for a
variety of reasons that have proven persistently difficult to fix well. C
and C++ are not unknown among mainframe applications, middleware, and
operating systems -- far from it -- but they're somewhat less common on
mainframes than in the IT world at large, in percentage terms. If you hold
that viewpoint -- and you probably are correct in that viewpoint -- then,
ceteris paribus, a more mainframe-oriented computing environment would be
inherently more secure. Hard core, mission critical, mixed workload, time
sharing environments -- the z/OS mainframe being the preeminent example --
also get some security help from the fact that they have to defend well
against "accidental invasions." These are not reasons to rest easy, though.
Healthy paranoia has merit.

--------------------------------------------------------------------------------------------------------
Timothy Sipples
IT Architect Executive, Industry Solutions, IBM z Systems, AP/GCG/MEA
E-Mail: [email protected]

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to