![[LWN Logo]](/images/lwn.banner.gif)
Date: Tue, 15 Dec 1998 11:50:35 -0500
From: srctran@world.std.com (Gregory Aharonian)
To: Jonathan Corbet <corbet@eklektix.com>
Subject: Re: IBM patent message
Sure - use the article. If you have room, here is some information on
my Internet Patent News Service, which is free, something I think will
be of increasing interest to the Linux world. Also attached is some
information on software patents, which you can send out as well.
Greg
INTERNET PATENT NEWS SERVICE
How to subscribe
The Internet Patent News Service is a mostly daily news service dealing
with information about the patenting world. Topics include announcement from
various Patent Offices around the world, stories about who is suing who,
interesting new patents and styles of patent drafting, statistics on issued
patents (with a focus on software patents), reviews of patent books and
computer programs, and biting commentary if I happen to be in a lousy mood
that morning.
Additionally, contributions are solicited from anyone with some patent
issue they have an opinion on, as well as explanations from patent lawyers
about some of the issues raised in the news items.
Also we have a Web site with a variety of files on patent issues
at: http://sunsite.unc.edu/patents/intropat.html which amongst other
things has the US Manual of Classification with patent title retreival by
patent class/subclass.
To subscribe to the daily news service, please send the message NEWS
to the following email address. Also, if you don't mind, please include some
information on what you do and how you might use this patent information.
Forward the request NEWS to:
patents@world.std.com
PLEASE INCLUDE YOUR FULL NAME AND AFFILIATION.
Please pass the word, especially to those not on the Internet or that
don't read USENET. Patent information is very valuable for finding out what
others are doing, for locating new technologies to license, and to measure
rates of progress in other fields.
For $40, I sell an MSDOS hypertext archive of past news stories,
currently over 400 taking up a few megabytes of space. To order, send
a check to the address below.
If you have any questions, please contact me at patents@world.std.com
and I'll get back to you.
Gregory Aharonian
Source Translation & Optimization
P.O. Box 404
Belmont, MA 02178
617-489-3727
patents@world.std.com
!19981018 17,500 software patents to issue in 1998
1998 Software patent statistics - Jan. to Sep.
Based on an analysis of 3336 software patents issued circa January/August
of 1998, I have put together the following statistics. Simply put, in 1998
and 1999 the PTO will issue 40,000 software patents, ten times the amount
issued six years earlier in 1992 and 1993, without ten times the resources.
In short -
- 1998 quicky stats: 17,500 software patents; IBM/Microsoft/Sun/HP
in top ten, 300 Internet specific patents (even more in 1999),
700 financial/business patents (today's Business Week 10/26 has
complaints from Wall Street)
- That a software patent issues means nothing about the patent's
novelty and/or obviousness - >50% of these patents are invalid
because >80% cite insufficient to zero non-patent prior art
- US software patent system (PTO, patent bar) has no credibility
beyond that of a statutorial registration system - Rule 56 is
a failure
- Large companies getting the bulk of software patents use them to
create barriers of entry for other big, and small, companies
- 40,000 patents means $50,000,000 in application issue fees, and
$40,000,000 in maintenance fees some years later - definitely
helps make the PTO a profit center
- The quality of issued software patents can be improved greatly
within the PTO's current powers and budgets - it refuses to do so
so it won't lose the above fees
Here are software patent statistics for the 1990s decade, with the 1996
to 1999 figures estimates:
1999 22500 ---|
1998 17500 -------- 40,000
1997 13000
1996 9000
1995 6100
1994 4500
1993 2400 -------- 4,000
1992 1600 ---|
1991 1500
1990 1300
Incredible growth with no relationship to the minimal revolutionary innovation
nature of the software industry. There will be 100,000+ active US software
patents in the year 2000. Typical current issues of the PTO Official Gazette
are listing 500+ new patents each week - the week of Sept. 1st saw 518 new
software patents. Get ready to be sued.
How well are these patents being examined? More specifically, how well
are these patents being searched to assess novelty and nonobviousness under
sections 102 and 103? In short, poorly. The PTO refuses to give examiners
enough time, money and resources to search. For yet another year, over 50%
of software patents cite NO non-patent prior art, the average patent citing
only 3.1 non-patent prior art items, which is woefully inadequate, since
there are dozens of highly relevant non-patent prior art items for most of
the patents being issued. And too many large companies are hiding behind
the impotence of Rule 56 to not do any searching themselves. Some numbers:
Number of non-patent Percentage of issued
prior art citations software patents
-------------------- --------------------
1998 1997
0 53 51
1 12 14
2-3 14 15
4-10 15 15
>10 6 5
Average number of non-patent prior art cited in 1998 - 3 articles
Average number cited ignoring outliers - 2 articles
Average number of prior US patents cited - 11 patents
Average number of prior foreign patents cited - 1 patent
Less than six percent of the patents cite even half of the relevant amount
of prior art. As I shall argue later this week, bibliographically this
means software patents should be presumed to be invalid. Patent 5,796,943
is a good example of the system not at worked - a smart card patent that
cites NO patent prior art and NO non-patent prior art.
Sadly, these numbers have changed little since I analyzed 1000 software
patents in the first few months of 1994. Thus since Compton's multimedia
patent forced the PTO pay attention to software patents, the PTO has made
absolutely no progress in improving the quality of issued patents, despite
the unsubstantiated comments to the differ made in public by the heads of
the various computing examining groups (2300/2400/2700). If anything, the
quality of patents has dropped substantially, given that there is no way any
government agency can increase its output TEN fold in six years while having
increased its process resources (examiners, tools) much less so.
So despite 4 years of gimmick fixes by the PTO, the 'wisdom' taught by
the Scatological Patent Institute, stick-heads-in-the-sand agendas of the
software committees of the AIPLA, and examiner access to the Internet (yes,
I know examiners are using Yahoo/Altavista/etc. - I don't for my busts),
little has changed in the quality of issued patents. Couple this with a PTO
management style that deliberately encourages invalid software patents to
issue, as one examiner recently lamented to me:
Greg:
One of the reasons so many "crappy" software patents issues is
that examiners are forced to do so. The PTO's production system
discourages CONTINUED searches is a really big problem, for 2 reasons:
1) upon 1st amendment, PTO management's strong encouragement to
either: issue or make final (do NOT continue searching)
2) in effect, actually taking AWAY counts for finding new art
with which to make a rejection (for continuing to search)
Why? We not only get no points for non-final-rejections,
but a non-final rejection also takes much longer to finally
go to a count (abandonment or issue)
So if they rewrite their claims to get around our first action, they
usually win, especially if they throw in additional claims to drag
things out. Thus, we might throw a couple of good punches, but
essentially we always are forced to quit the fight in the first round.
And applicants know this.
So patent applicants and their lawyers hide behind the impotence of Rule 56
to not send in any non-patent prior art, forcing the examiner to rely mostly
on patent prior art in one round of searching to decide 102 and 103. A joke
if it wasn't so unethical.
Who is getting the patents? For the most part, it is American and
Japanese companies and individuals:
United States 60%
Japan 25%
Europe 9%
Asia 3%
Others 3%
Europe has yet to appreciate the trade barrier potential of software
patents, and EC bureaucrats should be more willing to spend money to deal
with this problem. Currently they are dragging their feet, to the great
disadvantage of the European software industry. And at least for software,
EPO does just as bad an examination job as the PTO - the EPO's better
reputation cannot be extended to software patents (comparisons of PCT
search results shows little difference between EPO and PTO).
WHO'S GETTING THE PATENTS
Microsoft and Sun are firmly in the top 10, with Apple probably soon to
join them. MIT is the lone university with significant numbers of software
patents. IBM, still number one in (invalid) software patents, though its
percentage is dropping. For years, IBM had been getting about 9 percent of
the software patents, though in 1998 it is in the area of about 7 percent.
ATT dropped out of the top 10, with its spinoff Lucent taking its place.
Automobile companies continue their rise (Ford, Toyota, Honda).
Estimates of software patents by companies in 1998 (based on 3336 sample)
1200 IBM
360 Motorola
330 Fujitsu
330 Canon
310 Microsoft
300 Lucent / BellCore
280 NEC
260 Sun
260 HP
250 Sony
250 Hitachi
240 Xerox/Fuji Xerox
240 Mitsubishi
230 Intel
220 Toshiba
190 Apple Computer
160 Siemens Aktiengesellschaft
150 Samsung
150 Eastman Kodak
140 Telefonaktiebolaget LM Ericsson
130 Compaq/DEC
120 ATT
110 Nikon
110 Matsushita Electric
100 Seiko
100 Toyota
100 Sharp
90 Ricoh Corporation
90 Honda
80 Nokia Telecommunications
80 U.S. Phillips
80 General Electric
70 MCI Corporation
70 Bell Atlantic Network Services
60 Yamaha
60 Pitney Bowes
60 Olympus Optical
60 NCR
60 Mita Industrial
60 Ford
60 Brother Kogyo
50 Oracle
40 MIT
These companies mostly have a strategy of quantity over quality so that they
can play licensing games (often when IBM comes knocking - see the 3/17/97
article in Business Week). That some of the biggest also dominate the
Software Patent Institute makes it not surprising that the SPI is a fraud.
SOFTWARE PATENT CATEGORIES
What's being patented? Everything still for the most part. Networking
patents increasingly dominate, with automobile/vehicle patents continuing
their steady rise. Optics is also becoming popular, with the camera and
copier companies getting lots of software patents.
Estimates of software patents by categories in 1998 (based on 3336 sample)
3800 Networking / Communication
2200 Operating Systems
2000 Image Processing
1350 Vehicles
1350 (Graphical) User Interfaces
1200 Graphics
1200 Process Control
990 Signal Processing
980 Office Automation
940 Security
930 Engineering
840 Database
840 Optics
830 Medical
810 Numerical Analysis
710 Biological
700 Financial
650 Distributed Processing
580 Digital/Analog Circuit Design
500 Speech Recognition/Synthesis
510 Pattern Recognition
475 Computer Aided Software Engineering
470 Physics
410 Navigation
410 Word Processing
360 Entertainment / Multimedia
310 Compression
300 Internet
300 Object Oriented
290 Computer Aided Design
280 Chemical
250 Games / Gambling
240 Geophysical
220 Character Recognition
200 Neural Networks
190 Multiprocessing
190 Natural Language Processing
170 Manufacturing
160 Algorithms / Data Structures
150 Robotics
150 Artificial Intelligence
120 Agricultural
100 Music
95 Simulation
95 Parallel Processing
90 Fuzzy Logic
80 Genes/Genetic Analysis
75 Education
50 Virtual Reality
25 Vision
20 Spreadsheets
6 Quantum Computing
6 Java
6 Year 2000
Topics slipping in the ranks include music, education, word processing, and
artificial reflecting industry trends.
PTO EXAMINING GROUPS
Who is examining these patents? About two thirds are still being handled
by the Computing/Communications megagroup of Groups 2300/2400/2600/2700,
but other groups are handling one third of the software patents, a large
number for which some of these other groups are not qualified to examine
(worse by not having any software search resources directed their way).
Further I see some companies filing most software ideas written in such a
way that it is handled by the non-software/hardware groups.
PTO Groups
------------------------------------------------------
35% - Computing Megagroup 2700
12% - (was mostly hardware) 2300
8% - (was mostly software) 2400
66% 11% - (was telecommunications) 2600
------
34% 11% - Ind. Elect./Special 2100/2200
4% - Electrooptics 2500
10% - Mechanical 3100/3200/3400/3500/3600/3700
7% - Medical Devices 3300
2% - Chemical 1100/1200/1300/1500/1800
How about other countries? Based on filings, my rough estimate is that
there are twice as many Japanese software patents as there are US, though
the Japanese patents tend to be these short, low number of claims, patents.
My other guess is that there are about half as many European software
patents as US. China is a big unknown, with more patents than Japan.
All together, as we enter the next century, there should be about 300,000
active software patents around the world.
Greg Aharonian
Internet Patent News Service
(for info on free subscription, send 'help' to patents@world.std.com )
(for prior art search services info, send 'prior' to patents@world.std.com )