To: consultation@eurolinux.org
Subject: EuroPatent
X-Uri: http://www.chaos.org.uk/~eddy/
Reply-to: Edward Welbourne <eddy@chaos.org.uk>

To those involved in patent law,

It has come to my attention that moves are afoot to extend the laws on
patents to embrace various `soft' businesses.  I am a software engineer
by profession and, in my own time, write and publish web pages on a
variety of topics.  The proposed extension of patents would impose a
significant impediment on my work: indeed, I would argue for a reduction
in the scope of patents, not an expansion.

When I write a computer program, I analyse my customer's problem and
think about ways to solve it.  This naturally involves formulating
algorithms as I proceed with the work.  Since my customers' problems
usually suggest solutions, and I have a certain facility finding
solutions and designing algorithms, I almost invariably solve problems
without reference to external literature.  In this I am not unusual,
within my profession.

Meanwhile, the published literature (or, indeed, just that tiny portion
of it consisting of software patents issued in the USA) has grown to
such proportions that it would be unrealistic to expect me to read
enough of it to be sure of seeing all published material which bears on
my work.  If I attempted to keep up with the literature, I would have no
time left to earn a living: and would need to earn more than I do at
present simply to cover the cost of obtaining access to the literature.

Meanwhile, as the USA has learned to its cost, software patents would
impose a burden of review on the patent office which does not appear, on
the face of it, to be faithfully dischargeable.  Software is malleable to
a degree unparalleled in any other field: ideas prompted by solving
problems in one area of the field routinely have valuable applications
in others; transposing an idea from one branch of software to another is
obvious and far from novel; consequently, a examiner considering an
patent application in one field must needs have kept abreast of all
branches of software to be sure that the idea is novel and non-obvious.
I do not believe that, in practice, an examiner could hope to do this.

A large enough team of examiners would be able to read all of the
literature: one might then hope that an examiner could rely on her
colleagues to advise her on the novelty and obviousness of a given
patent application.  Likewise, a team of software engineers could share
the burden of reading patents and hope to collectively determine whether
each algorithm any of them used was infringing.  However, in both cases,
the hope is vain: the patent examiner or programmer could not reliably
notice the questions appropriate to ask of her colleagues, so would in
fact have to pass the patent application, or algorithm, around the
entire team for review.  The cost of reviewing a patent, or my use of an
algorithm, would become enormous: each member of the team would have to
spend a large proportion of her time discussing what she understands of
her share of the literature and reviewing the work of her colleagues; the
productivity of the team would be negligible.

Even were I to have read all the patents that have been filed, and yet
still, somehow, find time to practice my profession, I should be at a
loss to know which patents actually cover any particular algorithm I
might consider using - or, conversely, to know whether the algorithm I
have just thought up was `obvious' or constituted an example of genius
on my part.  It would appear that the only effective method of
determining such matters is to go to court and see who wins: given the
advantages available to those who can afford better lawyers, I have
trouble distinguishing such a system from a mediaeval `trial by combat'.
Surely, the law should enable me to know, before I embark on my labours,
whether I may realistically hope to enjoy the fruits thereof ?

To be able to practice my profession, in the presence of software
patents, I would need

  to learn patent law (a major disruption to my career: lawyers take
  half a decade or more in their training and are not fully competent
  until they have tested their knowledge by fighting law suits for some
  time)

  to read all the patents published which *might* impinge on my work
  (which does not appear to be a feasible project, by any stretch of the
  imagination)

  to review each algorithm I write not only for the present criteria (of
  which correctness and efficiency are relatively easily described to a
  layman, but by no means the whole story even now) but also for conflict
  with patents.

These would become professional duties - to neglect them would expose me
to risks of litigation.  Even were I to do these things, I would be
exposed to the risk of litigation by someone whose patent I, in good
faith, believed I was not infringing; and if they have better lawyers
than I can afford I should likely be obliged to pay them licence fees
rather than bankrupt myself with the legal fees attendant on fighting
the case.

Any sizable body of software involves many algorithms, interwoven in
many ways; and the interweaving would be as amenable to patenting as the
algorithms interwoven, since there is no clear distinction between
choosing algorithms and weaving them into a whole.  Consequently, it
appears likely that the cost of developing even quite a mundane software
package would be substantially inflated if the only added cost, due to
patents, was the need to license other folks' applicable patents.

In fact, of course, the time taken to discover which patents I should
licence would ensure that my productivity - and that of every software
professional - would plummet.  Globally, there is already a shortage of
competent software professionals: if legal burdens were to slow us all
to a fraction of our present productivity, that shortage would be
immensely exacerbated.  This would certainly exacerbate the increase in
the cost of developing software.

My only possible defence against the cost of licensing would be to take
out `defensive' patents in hopes of being able to negotiate
cross-licensing deals.  Even this would be of limited utility: a
competitor holding a patent that covers part of my business can
effectively exclude me from earning a living; and, in a modest time, one
may rely on the industry to become completely controlled by a network of
cross-licensed patents.  This would amount to a cartel: I believe it is
widely accepted that cartels can be relied on to be anti-competitive.
In any case, the structure of cross-licensing unavoidably raises an
immense barrier to entry which would prevent businesses from moving into
the sector; this would undeniably impede competition, regardless of the
general arguments.

The immense malleability of software would ensure that a patent-holder
would be able to claim, without provable bad faith, infringement by a
wide variety of software businesses.  Each patent-holder would thus be
able to hold many businesses to ransom, obliging them either to endure
the massive costs of a lawsuit or to pay license fees.  The body of
law-suits resulting would swamp the courts; each suit would be highly
contentious (hence expensive both to fight and to adjudicate); and those
of us still able to spend any time writing software would be deprived of
any guarantee of being able to reap the just rewards of our labour.

As it presently stands, being first to market with a software product
generally gives a software producer a major advantage over later
competitors, provided the product is well-integrated with
widely-available software packages.  However, this advantage may be
undermined in two very direct ways: first, the packages with which one
must integrate may have interfaces protected by licensing restrictions,
a peril which would be greatly reinforced by patents; second, someone
claiming that my product infringes their patent can (for the duration of
the law-suit needed to prove the product does not infringe) prevent me
from publishing - thereby keeping me out of the market while others
devise competing products, thus robbing me of the natural advantage of
being first to market.

Thus software patents would
  destroy competition in the software industry,
  vastly inflate the cost of software development,
  significantly reduce the productivity of the profession,
  deprive the software industry of any security of reward for labour,
  swamp the courts with an unending tide of expensive cases,
  deprive innovators of one of their major advantages and
  be impossible to examine rigorously.

The very existence of patents would impede various other business models
the innovator might otherwise exploit.  If I do some consultancy work
for a business and, in the process, come up with a novel and non-obvious
idea which expedites that work, I should like to be able to apply the
idea equally in subsequent contracts.  However, it may safely be
supposed that I do not come up with the idea until exposed to the
details of the task in hand, by which time I shall have signed a
contract with my customer.  In the presence of software patents, one may
be sure their lawyers have insisted on a clause giving them control over
any ideas I have during that work: I shall not be able to exploit that
idea in the service of other customers, even though I might confidently
say that I would have come up with the idea, had I not taken the first
contract, just as soon as I was exposed to the problem in hand.  I could
refuse to work under any contract which exposed me to this risk: but I
would almost certainly thereby limit my available employment
disastrously.

Having reviewed the principal, and least controversial, costs of
software patents, let me now consider the benefits.

There would be clear benefits to monopolists and cartels in the software
industry: they would be able to exclude newcomers from the industry with
even greater ease than at present; indeed, patents would practically
force the industry into a cross-licensing network which it would be hard
to distinguish from a cartel.  But I do not believe we can fairly regard
this as a benefit to the public in general, so let us ignore it.

There would be clear benefits to patent lawyers - the scope of their
business would be vastly increased.  First there is the simple matter of
adding software to that scope.  Second, the cases would involve very
large amounts of money, and be very hard to resolve, so patent lawyers
could anticipate high rates of pay for such cases.  Further, each
software patent would potentially apply to many branches of software,
greatly expanding the scope for litigation.

Indeed, once I had completed my training in patent law and caught up
with the corpus of patents issued and prior art published in the interim
(professionally necessary, as noted above), the balance of economic
incentives would certainly argue that I should practice patent law,
specialising in software.  If I returned to writing software, I would
have no guarantee of being able to earn money by doing so (or, at least,
of being able to keep that money once the patent lawyers came to call).
On the other hand, I could apply my talents to devising algorithms and,
without taking the trouble to implement them in a manner useful to any
customer, obtain patents on them.

Given my technical skills, I could readily detect anyone subsequently
employing `my' algorithms and take them to court.  Given my legal
training, and the same quick mind that enables me to write good
programs, I would have a realistic opportunity of obtaining significant
money from those actually providing software for real customers.  As for
me, so for all the brightest and best in my industry.  So there would be
a benefit for me: I would make more money (and never make any more
software).  However, the software industry would die: it would be unable
to make any money, and those capable of writing good software would have
a more lucrative profession available to us, so the quality of such
software as got written would plummet.

Again, though some would benefit under this heading, the public in general
would plainly lose by it, so we cannot count it.


So what of the one benefit I have always respected in patents ?
Namely, that when someone has a good idea, and can make more money out
of it if they keep it secret, they may well die while no-one else knows
that idea; by offering them, for a limited time, control over the use of
that idea we are able to give them an economic incentive to publish
their idea, so that they shall not take it with them to the grave.

I cannot see how this benefit can possibly be said to apply to the case
of software patents.  The benefit, on the face of it, is void if the
inventor can make no money from the idea without putting, in the hands
of the public, that which will reveal his idea to his peers.  In the
case of software, while the inventor may disguise the idea in various
ways, he cannot help but place a program in the hands of the public:
given enough time, the collective intelligence and ingenuity of the
world's software professionals shall discover his idea.

Furthermore, in practice, to make any money out of their idea, the
innovator must - in the modern business world - share it with their
business partners, their employees, their suppliers and, in most cases,
their customers: they may contractually bind these parties to secrecy,
but they will not take the idea to the grave when they die.

[Indeed, it is for this reason I would argue for a *narrowing* of the
scope of patent law: for the law to limit the public's freedom to make
good use of an innovation - not only to copy it but to extend it and
explore variations on it - it must show some benefit to the public
arising from this limitation.  As I see no certain gain to the public
but in the case where the innovation might be lost to the world once the
innovator dies, I would argue against the vast majority of modern
patents, from which the innovator could not possibly make any money
without disclosing it to others.]


Now, it is argued that the public gains one other way by patents:
namely, that by rewarding innovators we stimulate innovation; it is
argued that patents offer researchers some security of gaining a return
on the investment they make.  Yet in fact patents take away at least as
much, under this heading, as they give: when (as often happens) several
research groups come upon similar innovations at roughly the same time,
only the first group to apply for a patent on its work is able to
profit.  All others, far from gaining security of return on their
investment, are deprived of any return at all, save by grace and favour
of the first - who, as their competitor, is not apt to be generous in
this matter.  The patent-holder undeniably gains by this, but only to
the extent that the public and his competitors lose by it.

One might argue that, by rewarding the first to publish, we encourage
researchers to work faster.  Indeed, we do: but we move the target at
the same time - the incentive motivates pursuit of those parts of the
research which the patent lawyer needs, to draft the means to control
the new market, at the expense of research towards producing the best
product in that market.  The consciencious researcher who holds the
needs of the customer foremost in her mind is penalised for the sake of
the cynical exploiter of patents, who need not even take the trouble to
deliver a service to the public but yet profits by the conscientious
researcher's good work.

Furthermore, even when only one researcher pursues a theme, once that
researcher has obtained a patent on it, other researchers are thereby
subject to a major dis-incentive to research enhancements and extensions
of that innovation: for they can only profit by such work at the whim of
the holder of the patent.  As the field of software is characterised by
an endless race to improve on the work of one's competitors, which race
is the main-spring of innovation in my industry, patents would plainly
discourage innovation in the software industry.

Even without patents, there is ample incentive for businesses to pursue
research and to innovate: I have discussed the `first to market'
advantage above, and the natural and general advantages of having a
superior product may be added to this; we do not need to hobble our
industry to add a doubtful and unreliable extra incentive on top of
those which arise naturally.


Consequently, while I can see a whole horde of costs attendant on
patents, at least when they are applied to software, I can see no
benefit to the public, at large, save some speculative chimerae.

One might yet argue that, since the most powerful economy on the planet
has adopted this folly, we should follow in their footsteps.  However,
Europe is rapidly on the way to displacing the U.S.A. as the most
powerful economy: and may be expected to make even faster progress, as
their software industry falls into disrepair under the litigious burdens
of their folly, while ours finds its feet - as a visit to Cambridge will
plainly show you.

Let Europe's innovators but have the chance to compete with one another
for the European market (and, doubtless, impede the American industry
further by taking out patents in their jurisdiction) and we shall soon
enough become a power to be reckoned with - and a major source of tax
revenues to the governments of Europe.  We shall fare better in a free
market than in one burdened with patents - the last vestige of mediaeval
protectionism.

I urge you to reject the proposal for software patents.

Yours faithfully,

	Edward Welbourne.
--
http://www.chaos.org.uk/~eddy/