GREGORY AHARONIAN

SOURCE TRANSLATION AND OPTIMIZATION

MR. AHARONIAN:  I run a consulting service up in the Boston area, dealing
with software re_use technology transfer, and more recently software prior
art and software patenting.

And to summarize how I feel with what's been said today, I am for software
patents, for better prior art, open re_examination, and I'm against
prepublication.

No, I am not a patent lawyer, or have any software patents on file.

The chief asset of my consulting business is what I consider to be the
largest software prior art database in the country.  I have information on
over 15,000 government, corporate, and university research programs, 5,000
patents, and over 100,000 journal, article, technical reports and books and
other such things in which software technology might be described.  I
monthly monitor the output of about 150 corporate academic and government
research labs and about 250 journals.

The problem of software prior art is very, very nontrivial, and I think
that's why so many other government agencies have had problems over the
years dealing with the similar issue of how do you track all this country's
software technology.

I have many more comments on the art of software prior art, and I'd like to
speak about them in February hearings in which it's more appropriate.  What
I would like to speak about now is kind of observations on having examined
so much software over the years on a few of the issues that have been
brought up today.

One, is I think it's going to be very difficult to change the rules to deal
with software patents.  There's already a current set of statutory
guidelines that are pretty well_reasoned, pretty consistent, certainly
comprehensive, if you read all the court cases.  But I don't feel that
they're working well, for the following reasons: I've examined over 5,000
software patents as part of my technology transfer business, and I've seen a
lot of very trivial software concepts actually getting a patent.  I've seen
many software patents with very broad claims, I've seen things that are as
close to a pure math algorithm as possible with maybe one claim in there for
a piece of hardware, and I see more and more business practices being
awarded software patents.  And in some cases, I see pure source code being
patented.  In 1992, the Air Force got the patent for the difference in
source code between two versions of a public domain program at Ames.  And I
consider that to be fairly trivial.

If you try to come up with more rules to guide the process of awarding
software patents, I think you're just going to come up with more ways for
patent lawyers to get around the rules.  It's tough to treat software any
differently than any other technology, and I think you have a lot of
problems.

So in general, based on just examining a ton of software in the past 10
years, I'm not sure there's much anyone can do in terms of trying to come up
with more rules, it's very difficult.

And on the practical side of the patents I've examined to date, I could
probably successfully challenge 25 percent of them on software prior art and
a few related issues.  There's a lot getting through that should not be
awarded.

Of course, my phone is not off the hook, asking for my services to challenge
software patents, so I still don't think to date it's a big problem.  I
think it's being exaggerated because of stunts like the Compton incident,
which is great for PR, but not much more.

The second issue I would like to address is that of what one of the earlier
speakers, Robert Sterne, testified that there's really no difference any
more between hardware and software, and that if you try to change the
software patent rules in isolation without treating the hardware patent
rules, you're not going to do anything.  You're going to leave a big
loophole for people to get around the hardware rules by just doing things
over on the __ or getting around the software rules by doing things over in
the hardware world.  It's even happening today.

Existing technology now in the market where I can take a circuit schematic,
which anyone would consider a piece of hardware for the most part,
automatically convert it into a computer representation language and then
convert it once more into an algorithm which I think most people consider to
be software.

Similarly, there's other technology out there that let's me take an
algorithm written to a traditional language like Pascal or C, covert it into
another intermediate language and feed that into a hardware design tool, and
get out an integrated circuit.  So here I'm starting out in the software
world and ending up in a hardware world.  To me, it becomes impossible to
find what software or hardware is.

And quite recently, in fact, last week a company in Germany announced a tool
that integrates __ and that's the overhead I have up there __ computer_aided
software engineering tools which is the domain of the software world with
hardware design tools, so that within one tool set I can type algorithms,
draw circuits and go back and forth and not really care at all, the computer
will take it all and account for me.

And with such a tool I can design a new device, and with a cleverly_drafted
set of claims, where I have a broad independent claim that talks about
systems and methods and things of that nature, dependent claims somewhere
that actually mention hardware and software, I can protect infringements in
both worlds with one patent.  So to make these distinctions between hardware
and software, I think is a mistake.

And in fact, there are a few of these design tools in which it should be
possible within a year or two to not only allow the user of the tool to
generate either an integrated circuit or a computer program, but at the same
time a patent application.  And if you want to have a million software
engineers and electronics designers having an automated patent generation
tool on their hands, it's going to become a possibility quite soon.  And I'd
hate to think of the headaches you're going to have then.

One other thing, the equivalence of hardware and software also complicates
the issue of software prior art, because if these mappings are true and you
can go back and forth between the two domains of hardware and software, and
conceivable someone's circuit schematic somewhere could serve as software
prior art in another case.

And I track both software and hardware, so it's no big difference to me, but
if the Patent Office intends to seriously treat the problem of prior art, it
will not be able to do software prior art separate from hardware prior art. 
It is all one field of computing devices.

Now, one thing I would like to suggest is currently I publish over the
Internet a news service dealing with patent information.  Each week I mail
out to about a thousand sites on the Internet the titles and numbers of the
latest patents coming out of the Patent Office.  In Boston there is an APS
terminal that I can use for free, which is very nice.  And just once a week
I go down there, and dump off a couple files worth of data, broadcast it out
over the Internet.  It's a very well_received service, I offer it for free
since it doesn't take too much of my time.  And the main demand is for
people who are trying to find out more information on software patents.

At best, if someone actually cared to make an effort to find out what was
being patented in the software world, you would go to a local patent
repository and use one of the CD ROMs, (Casus Bib) or something, and those
tend to be five or six months out of date at the typical repository.  And
people tend to want more recent information.

So the various calls to get the APS system on the Internet will be very
well_received in the Internet world, and there are many out there glad to
help out with the process.

So In general, I don't think there's much you can do to change the rules of
dealing with software in isolation.  The current rules are a good set of
rules.  I'm not a patent lawyer, but rules are rules.  I don't think you're
going to be able to do too much better.  You can change some of the
procedures, and many people are calling for that.  But dealing solely with
statutory type things, I don't think it's going to have much of an effect.

And certainly, if you had seen a lot of the software patents I've examined
over the past four or five years, it's hard not to conclude that they just
don't work.  I think the open re_examination process will help, but I think
you're grossly underestimating the amount of paperwork and headaches that
that's going to entail.

If everyone in Internet, at the request of someone, decides to forward their
pet document to the Patent Office, I mean, you'll get a million people
sending in something that pertains to one particular patent.  You could fill
up this hall many times over for each case.

That's all that I have to say.

COMMISSIONER LEHMAN:  Thank you.

One thing actually I had meant to ask one of the other witnesses, but since
you indicate that you see a lot of patents issued that you're quite
convinced would not meet the nonobviousness test __ our decision to
reexamine the Compton case was a very unusual one, but the Commissioner does
have the power to order re_examination himself.  What would you think of as
one method of attempting to clean the files of allegedly nonpatentable
incorrectly issued patents if we had some kind of a program where maybe we
work with people like you and attempted to identify some of these patents
that we issued, we ordered our own re_examination?

(applause)

MR. AHARONIAN:  Well, I think it's a great idea, and certainly would love to
bid on a contract to do that.  There are two practical problems I see with
that.

The first is, I'm actually an inventor in the world of electric power
equipment.  That's how I got into a lot of this patenting stuff.  I don't
think the problems of prior art in the software world are any worse off than
in some of the other fields.

For example, many of the high temperature superconducting patents for
devices being issued today would be invalidated by low temperature
superconducting device patent applications dating back to the '30s and '40s. 
Most of those patents no one knows about anymore.  You really can't get at
them through APS or any of the CD ROMS or anything else.  They're literally
lost unless you go look for them.  But they couldn't validate many of the
new high temperature superconductor device applications.  So I could argue
that if you're going to consider doing that in the world of software, you
could do it in all the fields because they all have that same problem.  And
most information dating back before the early '60s or so in any field is
literally off the abstracting services of anyone, and it's a big problem.

The second problem is, while I've been out here I've been kind of working
the venture capital circuit to see if I could actually raise some funding to
actually start a business for providing software prior art and services. 
I'm going around looking for $10 million, it's a very expensive process to
keep track of everything.  And I happen to be good at it, I mean if anyone
else was going to do it, I would say it would cost 20 million or more.  I've
been doing it for 10 years, I have 10 years of leg_work out of the way. 
It's a very expensive thing.

I know the Defense Department spends $10 million a year just trying to track
all of its software, and they haven't had much luck.  The DOE, NASA, all the
agencies have not had much luck.  It's a very difficult process of tracking
it all.

And I'll give you a case that will explain why it will be difficult when the
re_examination process opens up: In the signal processing world there's a
technique called a "Fast (40 A) Transform, it is used all over the place. 
In my life I've seen 200 different implementations of this one algorithm,
and as an algorithm it's not very complex to begin with, three or four
indented loops where some math goes on.  When I do my service, I have to
look at these 200 algorithms and figure out which four or five I'm going to
include in my database.  To make those type of decisions in all the
different aspects of software is very complicated.  You need the type of
person who's not trained anywhere.  I mean, I didn't go to school to learn
this, I had to look at the stuff over the years.  So the types of decisions
that have to be made in these re_examination processes are very complex,
require tons of data, and I'm not sure it even can be done, but I think
you're going to try your best anyway.

COMMISSIONER LEHMAN:  Thank you very much.Next, I'd like to call George
Cole.

Back to The San Jose Index

Forward to George Cole