NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/org/eff/talk/6935

Internet Message Format | 1992-11-11 | 6.7 KB

Xref: sparky comp.org.eff.talk:6935 misc.int-property:1424 alt.suit.att-bsdi:611 comp.unix.bsd:8745
Path: sparky!uunet!gumby!yale!yale.edu!think.com!barmar
From: [e-mail] (Barry Margolin)
Newsgroups: comp.org.eff.talk,misc.int-property,alt.suit.att-bsdi,comp.unix.bsd
Subject: Re: Patents:  What they are.  What they aren't.  Other factors.
Date: 11 Nov 1992 21:39:44 GMT
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 112
Distribution: usa
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]> <[e-mail]>
NNTP-Posting-Host: telecaster.think.com

In article <[e-mail]> [e-mail] (Scott Mcgregor) writes:
>In article <[e-mail]> [e-mail] (Greg Hennessy) writes:
>>I find this to be a strawman arguement. I have *NEVER* heard anyone
>>claim that patends cover what you can think about.
>
>It has been my interpretation that the "mental process" test is
>PRECISELY concerned with what you can think about, and not what you
>DO.

No.  The "concept of public-key encryption" is not the same thing as
"performing public-key encryption in your head".  Berstein's "mental
process" test has nothing to do with restricting what you can think about.
It is a litmus test for distinguishing pure algorithms (which should be
unpatentable) from physical processes (which are).  No one has ever
suggested that it be used the other way, i.e. that you no longer be able to
perform a calculation in your head if it's equivalent to a patented
process.

Perhaps instead of referring to "the concept of public-key encryption" we
should refer to "the field of public-key encryption".  I hesitate to use
the term "process", though, since I think this is too general a term to
deserve such a designation.  However, if the patent does indeed cover all
public-key encryption techniques, perhaps the PTO disagrees.

>Go back in time. Before RSA, ALL encryption was based upon
>PRIVATE keys.  The literature taught that it had to be that way and
>that key protection was very important. RSA was the first PUBLIC Key
>system. It broke the mold. It showed that regardless of whether the
>underlying algorithm was based upon multiplication of large primes, or
>ANY OTHER "one way trap door" function that PUBLIC key systems could
>be built and be made arbitrarily secure.  The innovation wasn't so
>much the use of multiplication of large primes, but of using PUBLIC
>keys.

That makes no sense.  That suggests that someone could patent public key
encryption without even specifying an algorithm that implements it!  The
innovation isn't the general idea of public key encryption, but the
discovery of an actual PKE algorithm.  Assuming algorithms could be
patented, I would grant them the right to patent RSA, but not the entire
public-key encryption approach.

Your view suggests that there could actually be a hierarchy of patents.
First someone invents an encryption process; he patents that process and
claims all of encryption (although his process happens only to be
symmetric).  Then someone invents an assymetric encryption process; first
he has to license encryption from the first inventor, and then he patents
the asymmetric nature (althouh his process only uses a particular
asymmetric algorithm).  Then someone invents a different asymmetric
process; in order to patent this, he must first license encryption from the
first guy and the asymmetric approach from the second one.  Is this really
how things are supposed to work?

If so, it seems to contradict things that have been said in other postings.
Someone (I don't know if it was you) suggested that part of the benefit of
patents is that it spurs invention of alternative approaches by people who
would like to find a technique that doesn't infringe someone else's patent.
But if someone can patent the entire field of asymmetric encryption,
there's little point in trying to invent an alternative PKE process to RSA,
since any PKE process would infringe by definition.  The entire field of
endeavor is locked up in this patent.

>>Thus, "cooking" is patented, as I use the terms.
>
>If I were to use the analogy, I would put it this way:  "For
>countless centuries, man has used flame to cook food: wood, gas, coals
>have all been used as fuel.  Now, RSA has invented a new stove. It
>cooks without flame, using electric resistance. They've been granted a patent
>on cooking using electricity.  It covers any materials that the "burner"
>is made of. It covers any shape electric "burner".  But it doesn't cover any of
>the pre-existing flame systems. You can still buy and build a gas
>oven. You can create another new flameless system using microwaves.

But it seems more like RSA has patented "cooking with electricity", even
though they may only have invented "cooking with an electric burner".  And
they're claiming that their more general patent covers cooking with
microwaves, since they're generated with electricity.  Such broad
applicability is preposterous.  People shouldn't be able to patent more
than they've invented.  The doctrine of equivalents should suffice to cover
insignificant variations.

Part of the problem is that it is almost always in the inventor's best
interests to claim as much as possible.  Some of the claims may be
indefensible (many claims of 4,555,775, Pike's "backing store" patent,
imply that he invented the windowed graphics terminal, even though they
were in use in universities and research labs several years before he
applied for the patent), and the patent holder may not plan on trying to
enforce those claims, but they'll probably deter some competitors from
infringing (would *you* like to go against AT&T in a patent suit, even if
you were sure the patent was invalid?).  I think something needs to be done
to prevent overly broad patent claims.

>Some company's license and market while others produce.  A local mom
>and pop grocery store doesn't do research to find out what people
>would like in a new breakfast cereal.

A grocery store doesn't charge you for shoplifting bread if you look at the
ingredient list on the package and then bake it yourself.

>Licensing companies don't research, develop or manufacture the
>products they license.  They may not even ship product.  But they
>provide a place for the public to acquire product.

Do people actually "acquire product" from licensing companies?  How many of
the people who've been approached about the XOR patent acquired anything
from that licensing company?  Regarding RSA, I believe most of us learned
the algorithm from any of hundreds of technical papers that have been
presented on the topic.
--
Barry Margolin
System Manager, Thinking Machines Corp.

[e-mail]          {uunet,harvard}!think!barmar