I thought this critique from a patent perspective of Samuelson et al's software protection proposal would be of interest to cni-copyright subscribers.
Norman Siebrasse
<siebrass[_at_]unb.ca>
!19960422 Critique of Samuelson and Kapor's new software protection scheme
This Friday there is an IP symposium at UCal Berkeley, a good deal of which deals with software patent issues. Since IBM and a few peers are involved with the symposium, it is not surprising that once again I was barred from presenting much of the data I have been collecting on software patents and patenting practices. Gee, banned at antiestablishment Berkeley in April, banned at establishment AIPLA in May. I feel so looooooooonely :-)
To illustrate why the type of data I deal with should be presented at such symposiums, I thought I would critique a recent paper partly written by one of the participants (PS) at the UCal Berkeley symposium. The paper is titled "A new view of intellectual property and software", written by Pamela Samuelson (Cornell), Mitch Kapor (MIT/ex-Lotus), Randall Davis (MIT) and Jerome Reichman (Vanderbilt). The article appears in the March 1996 issue of Communications of the ACM, page 21. They unsuccessfully attempt to propose a new way to protect software (as opposed to the current use of patents, copyrights and trade secrets).
The first problem is that they unrealistically characterize the types of software that need to be protected:
"In talking about the software marketplace, we are focusing
primarily on mass market software, which basically means
personal computing software (though we believe much of what
we offer carries over to other software markets). Where
product volume is small enough (e.g. mainframe software) to
permit it, individual sales contracts can be negotiated to
meet the individual needs of buyer and seller. When products
are sold in the tens of thousands to millions of units, only
standard, agreed-on rules are feasible, rules of the sort
provided by intellectual property law".
Had the authors bothered to examine the thousands of software patents issued
every few months, they would find that their characterization does not
generalize. People are seeking intellectual property protection for software
concepts that span market sizes and volumes of all sorts. Indeed 90% of
software patents are for software that is explicitly never sold commercially,
more typically being embedded into something else sold (cars, medical
equipment, imaging systems, telecommunications networks, etc.).
Next, they are overly pessimistic, even more pessimistic than me:
"Nevertheless, we argue with some vigor that while the changes
may be small, they are necessary; the existing system cannot
work in the long term".
They also fall into the trap that plagues that recently issued PTO's software patenting guidelines - asserting that "software" is a distinct concept:
"We suggest the problem is more fundamental: existing legal
concepts frame the issues badly when it comes to software,
and additional experience will only produce more thrashing".
Thanks to the activities of the design engineering world, hardware and
software are increasingly becoming arbitrary definitions. Some codesign
tools now coming on to the market allow you to describe, simulate and
optimize a system behaviorally, and then when you are satisfied, push a
button and have the computer tell you which part should be implemented in
hardware (generating the corresponding circuits) and which part should be
implemented in software (generating the corresponding code). From a legal
point of view, all that is important (for example, for Doctrine of Equivalents
arguments) is the behavioral description, which is neither hardware or
software. It wouldn't take too much to get these tools to also tell you
which parts of the designed system should be protected by copyright and which
by patents (and generate the necessary applications to some extent). As
these tools become more powerful, and if the legal world bothers to understand
what they mean, I see less thrashing. But this article for the most part
justs talks about "software", framing the issue just as badly as they claim
existing legal concepts do. It's funny, but at least two of the author's
institutions (MIT and Cornell) do some pretty good research in codesign.
The article then has a section titled "What's important about software", where they try to characterize software. For the most part, I am not sure what relevance these characterizations have (for example, a well written broad patent claim can protect a program metaphor), except that two of them are historically incorrect.
First, the argue that "Progress in software is typically innovative, not inventive":
"The character and pattern of technical advance is not
uniform across fields. Some fields are characterized
by discrete inventions, such as pharmacueticals, where
the discovery and use of one drug general proceeds
independent of others. Other fields, such as aviation
and automobiles, involve complex systems of interacting
components, and are as a result characterized by cumulative
advances in which new developments build on features of
the existing technology. Software involves complex systems
and is fundamentally cumulative. Advances in software are,
moreover, more often innovative. Improvements are incremental
rather than inventive in the patent sense of obviousness.
Few of the daily valuable improvements in programs pass the
relatively high threshold of inventiveness demanded of patents.
Again, the authors obviously haven't examined too many issued patents, because what they say about software applies for all fields of technology, including pharmaceuticals. Certainly all of the technologies in the Mechanical and Electronic section of the PTO Official Gazette are like software and hardware - incremental rather than innovative. Not that there isn't innovation occasionally - but so too for software. Even for chemical, much of it is not only incremental (700 new chemical patents each week for a decade with few changes in what are the hot drugs, speaks of incremental), but increasingly "use of one drug general proceeds independent of others" is not done. In any event, any assertion of software's uniqueness (and therefore needing of different laws) based on the above characterization is historically incorrect. Developments in other fields are as "boring" as those in software. Just look at the hundreds of thousands of new patents and journals articles each year.
Second, they, like most other writers and commentators, are historically wrong when they interpret the history of Xerox/Alto versus Jobs/Apple. In a section "Know-how on or near the face is vulnerable to copying", the cite a misleading industry myth:
"Perhaps the most famous example is the visit of Steve Jobs
to the Xerox Palo Alto Research Laboratory in 1979, for a
tour that included a demonstration of Xerox's Alto computer
and its graphical user interface (GUI). The visit resulted
directly in the use of the GUI in the Lisa and Macintosh.
Referring to the GUI, the head of the Xerox Lab subsequently
recalled: "To allow Jobs to see the power of the system ...
was a dumb thing to do ... Once he saw it, the damage was
done; he just had to know that it was doable". The know-how
was all there on the face. Know-how on or near the face is
immediately accessible to any observer and hence cannot be
kept secret.
Around the time of Jobs' infamous visit, Xerox was ordered by the Federal Trade Commission, under a CONSENT ORDER TO CEASE AND DESIST dated July 29, 1975, to make all of its copier patents available to its competitors royalty free (or for some economically pointless royalty). If you check out the November 4, 1980 issue of the PTO Official Gazette, there are 93 pages listing about 6000 patents that Xerox was forced to give up rights to. Xerox thinking then became "what's the point of patenting if the patent rights can be so denied", which lead to layoffs of lots of Xerox's patent lawyers and the general corporate message "don't get Xerox involved with intellectual property litigation", since the corporate management had had to spend much time dealing with the FTC action.
It isn't surprising then that in the late 1970's, Xerox employees were lax or unprepared in dealing with closely protecting Xerox intellectual property like the GUI interface. Why bother to protect the apparently unprotectable? So sure, Jobs did a great job of taking the GUI concept and marketing it through the Apple Computers (which Xerox could have done independent of the patent status of its inventions). But the authors are incorrect in using this historical incident as an example of some weakness in the current IP scheme for software. Mention Xerox and Jobs without mentioning the background FTC action makes for a cute story, but nothing else. A lot of people do this about Xerox and Jobs and they are all historically naive.
The authors then go on examine intellectual property laws in light of software, arguing that there is a mismatch between the protections offered by copright, patents and trade secrets:
"The problem is fundamental: intellectual property law gives
us three traditional mechanisms, yet the value in software
is not well protected by any of them".
Well, other than the problem of software prior art, and the troubles caused
by lots of bad software patents, I haven't seen too many in industry
complaining that IP protection for their software assets is worse than for
their other technologies. People infringe software patents/copyrights or
steal software trade secrets, but most incidents in the past have been
handled by the legal system without too many complaints - there are winners
and losers, but that happens everywhere. Sure some aspects need adjusting
(like closed patent reexaminations), but the bills are always being
introduced in Congress, some of which pass, to fix this. Even assuming
software is distinct enough to be compared, I do not see many people publicly
complaining that they can't protect their software assets. If Apple Computer
was upset about losing to Microsoft, that is not the fault of current IP laws,
they lost because of the historical context in which GUIs were developed
(look at the tables in the Patent Quarterly for this case - a nice example
of a prior art search).
They then argue positively that "Should software be treated differently"? For a variety of reasons, I do not think that software should be treated differently, because of things like hardware/software codesign, and general automata theory, I do not think that you could ever come up with a specific legal definition of software. Software is like pornography - I know what it is, just don't ask me to define it (making pornographic software utterly undefinable :-).
They sum up their arguments in this section as follows:
"As a consequence of all this, software is far more vulnerable
to a phenomenon we call "comparatively trivial acquisition
of behavioral equivalence", the speedy appearance of cheaper
products with equivalent behavior. Triviality of acquisition
matters because trivial acquisition destroys lead time,
thereby destroying the innovator's chance to recoup R&D
investment. Finally, we say "comparatively trivial" because
the second-comer's efforts may be far from trivial considering
on its own terms (e.g., Paperback's efforts in cloning 1-2-3:.
While I am jealous of the authors for coining that phrase, and agree that software products are susceptable to this phenomenon, I am not sure how microeconomically meaningful their argument is (and keep in mind changes to intellectual property laws are national in scope and thus of are interest to the country only at the microeconomic level). Was Paperback a threat to Lotus, or did it survive by feeding at the edges of Lotus' domain? If Lotus had had patent protection for their spreadsheet, would Paperback and any other competitors been able to be as successful (certainly with the monies Lotus had, with a patent, it could have tied up Paperback in court for enough years to reduce lost sales)? After all, Lotus' first patent issued in 1988, 4,788,538, "Method and apparatus for determining boundaries of graphic regions", years after 1-2-3 hit the market, and seems to indicate that Lotus' hadn't considered seeking patent protection at the outset.
Afterall, many of its peers (Microsoft, etc) weren't seeking patent protection while they built large companies. In fact, you would think that the authors, based on their histories and current academic status, would have thought to conclude at one point is that one solution that is very needed is that intellectually property concepts should be part of at least the gradute computer science curriculum - then maybe there would be less confusion. I'd be more than glad to give a talk at MIT - I use Barker, Hayden and Sloan libraries enough :-) For a bunch of academics to not pay attention to the IP shortcomings in the education system is disappointing.
I don't buy their arguments that some lack of software protection is a great hindrance to being a successful software company. A good product with a good marketing plan gives you a few years head start, and with good service and occasional improvements, you build customer loyalty for which some cheapy competitor is not going to make a big dent in sales. Afterall, the free GNU compilers have not shut down the compiler industry. On the other hand, a different way of doing the same thing (Lotus Notes versus intranet Web pages and USENET) is not going to be anymore protectable under a new IP regime than it is under the old regime.
They end the article with their suggestions for a new intellectual property protection scheme for software, which to me sounds like taking the existing laws and throwing them into a blender:
The first is copyrights, the third is patents, so what's the big difference? The second reduces to socialism (because someone [like the government] has to define, monitor and enforce this "market preserving period"), and as a legal change would never get through Congress (which is in this smaller-is- better thinking about government), assuming anyone could do a decent historical microeconomic analysis to figure out what a "fair" market preserving period would be (for example, an algorithm that can predict stock and futures market movements - what's its' value, and how long and how much should the inventor profit?). The fourth as a legal change would also never get through Congress (off-the-shelf liability principles - after the lawyers stop laughing they would lobby this to death), assuming again that anyone could do a decent historical microeconomic analysis to figure out what would be the terms of this off-the-sheld liability and standard licenses.
Additionally, the author's failing to treat hardware in their proposed plans weakens their case. If software protection laws are changed independently of hardware (which stays the same), then someone could protect a software idea by applying for a mostly hardware patent with a broad, Turing like independent claim, with a variety of hardware/software dependent claims, and then circumvent the new software rules by using the Doctrine of Equivalents to get longer software protection than the new software-specific rules allow. Everyone trying to treat software independent of hardware (like "lets get rid of software patents", or "lets change the software patenting rules") are wasting their time. But once you throw in hardware into the discussion, the reasons to change the rules become much less compelling than they already are (while increasing the complexity of the prior art problem).
Sure there are problems with current intellectual property systems. But other than are a few easily treatable but gross deficiencies (i.e. software prior art), the system seems to be mostly working, and certainly not worth scraping for a vaguely defined new system that the authors proposed, especially if the new system requires economically questionable concepts like "market preserving period" and "software metabolic rate", and especially if the authors base their ideas on extremely limited analysis of the history of software, and limited use of the demographics of software patents and copyrights.
Regretably for others, apparently Berkeley (and more importantly, some of the et.al.'s participating) isn't very interested in having this week's conference attendees be provided with much of the history of software and demographics of software intellectual property practices. Oh well, there is a good Middle Eastern cafe just off campus on Durant - if your head won't be filled, at least your stomach can be.
Greg Aharonian Received on Tue Apr 23 1996 - 16:23:48 GMT
This archive was generated by hypermail 2.2.0 : Mon Mar 26 2007 - 00:35:20 GMT