On Fri, 2 Apr 2004, Dan Kolis wrote:

> Dean Anderson said, and is ">"
>   >While finding prior art is hard problem in any field, it would be helpful
>   >if the Patent Office hired more experts in the fields that they offer
>   >patents in, and in particular, more computer scientists.  
> 
> Dan says:
> In the above, a chemist would substitute "Chemist" for computer scientist, a
> Mechanical engineer the same. 

Well, the PTO actually hires Chemists and Mechanical Engineers to review
patents in those fields. But it does not hire Computer Scientists.  At
least, it didn't. I don't think anyone's checked in a few years, but we
also haven't been told that this has changed.

> Obviously, the patent inspectors know there is probably public domain
> material, but it doesn't seem like they have a reasonable access to it.

One might think so. It is more likely a result of not having people
trained in computer science analyzing software patents. Perhaps having too
much work to do.

> The review process looks degrading to the patent inspector if the
> applicant goes over their head in their internal appeals process.  So
> its easier to grant it.

I don't think this the case. One nearly always has a recourse to challenge
a government decision. There is nothing degrading about doing so.  I think
most government bureaucrats try to do a good job, even if its in a field
they aren't trained for.  But their lack of training in the field is a
policy problem.  It doesn't mean there is anything inherently wrong with
the patent system.

Human failures sometimes mean there is something wrong with the system. It
could be the case that the system promotes human failures, like the Airbus
autopilot user interface that led pilots to mistakely enter 3500 feet per
minute descent rate instead of 3.5 degrees down decent. This led to at
least one crash and several near accidents. The system was fixed, so that
humans were less likely to fail. On the other hand, the aircraft that hit
the WTC performed perfectly. There was nothing wrong with the airplanes.

Similarly, if you have monkeys instead of railroad engineers, it doesn't
mean the concept of trains and railroads is flawed. It means that the
policy of hiring monkeys instead of trained professionals is flawed.  And
by "monkeys", I don't mean to be perjorative. Albert Einstein wouldn't be
safe driving a train, even though he wrote a lot about train cars and
railroad embankments. Its a matter of training and knowledge base.

In our instant case, while the policies are harmful and the computer
science training insufficient, this is a relatively minor problem. It is
the system itself that is flawed.  We do not want to focus on the
policies, except to say that they are relatively unimportant, though they
do create a negative.  Even if the policies were changed, it is ultimately
hard to avoid the problem of bad patents entirely and I don't expect
that's possible.

There have been people who have filed purposefully embarrassing patents,
but I don't encourage this. In fact, I strongly discourage it. Sabotaging
the system never helps change it.  In fact it hurts, because then the
proponents can claim that its problems are due to the sabotage, not to
anything inherently wrong with the system.

> Dean says:
>   >Such patents as this are clearly mistakes, and are frequently overturned
>   >on review. 
> 
> Dan says:
> As a reality check I just walked over to a desk here and touched an object
> recently contested in a Patent battle. *any* trip to a flea market could
> fill a pickup truck of goods built before 1940 that show the patent is at
> least partially invalid. (some claims... stink). But, the Grantee won. Why?
> Its cheaper to be right and lose, then pay a license fee... than be right
> and win.

True, "The right thing" doesn't always prevail. But I think it usually
prevails.  But these sort of mistakes are administrative--there are human
failings in _any_ system.  In principle, one can't criticize 'the system'
for human failing, because only the humans failed.  There is a difference
between the system and the operation of the system.  We can and do
criticize the operators for having too many human failings.  This is a
reason to change the policies of the operators, but not a reason to change
the system.  We seek primarily to change the system, and want to focus on
that.

> Dean says:
>   >if you file right before product
>   >release, and that product catches on. 18 months is a long time for
>   >software. 30 months, and you are into lifecycle maturity. You've already
>   >made commitments to using the software.  Now you have to pay whatever they
>   >want to charge.  If the patent is solid, there is no way out, not even for 
>   >OJ or MJ, or BG for that matter.
> 
> Dan says:
> The general principle is built on pain and suffering. The trick is, (like
> Sam Ting said about how to win Nobel prizes: "I think you should be first,
> and be right"). "Interference" is super complicated, when they actually
> overlap like you described. I think if I remember correctly, it happens in
> about 5% of the cases, so its a lot. No doubt, its a pipeline, so one begins
> to wonder what public interest is served with long delays... I can't think
> of any.

The delay isn't an "intentional" delay, in the sense that they just sit on
it for 18 months. The delay is to process the paperwork and review the
patent.  The delay is also due to a backlog of work.  Its no different (in
principle) than if you file for a building permit: you don't get it same
day. It takes a bit. If there is a backlog, it takes longer.  At the PTO,
it generally takes 18 months, but occasionally takes 5 years. (There are
exceptions having to do with classified technology that can take much more
time--effectively unlimited time, but I'll ignore those)

> A usual way to deal with the reality is to use the granted one as a
> citation, make a trivial improvement, and now you have reset the clock 2
> years or so. Of course, if there are real damages, all this is retroactive,
> plus often a bonus multiplier of 1:3 for being a evil-doer, so its a
> financial burden.

True. Though anyone can do this. If you want to screw your competitor, or
buy a company at a discount, you can file for the improvement and prevent
them from using it.  I suppose it is a financial burden, or an income
source for patent attorneys.  The counter argument is that people were
still spurred to invent new things, and to make improvements, and of
course, inventors are motivated to work harder and file sooner for the
enhancements to prevent competitors from doing so. So in fact, the patent
system is "working" in this case and promoting progress. This is just sort
of the "Stick"  side of it.  "Innovate or Die" makes progress. You may not
like that, but it isn't a strong argument to change the patent system.

> Dean says:
>   >But anything halfway novel, and new. Well, that is another 
>   >story:  Patent it or someone else will.
> 
> Dan says:
> What is "Obvious to one practised in the art"; (which is the US PTO test for
> novelty). Our patent attorney says if it takes more than 45 seconds for the
> dumbest person actually employed in that field to figure it out, then its
> not "obvious". 

Well, that's not the test that the patent office uses, thankfully.

Also, keep in mind that my comment is about the first-to-file system,
which hasn't been passed into law yet, in the US.  Though, it is fair to
say that even the current system strongly encourages patent filing over
public domain publishing.  For example, if someone else gets a similar
patent, and you have published in the public domain, your only recourse is
to challenge the validity of the patent, and you have very little leverage
with which to negotiate a cross license. If you had instead filed for a
patent, you would have some leverage to counter sue them for infringement,
and probably obtain a cross license.  This creates a strong incentive
against putting things in the public domain.

> Who know for sure, but I doubt this was the intention 300 years ago when
> this concept emerged.

Actually, the intention (more than 300 years ago) was to create arbitrary
monopolies. Patents were gifts of the King--you got to be the exclusive
maker of shoes, for example. It was a way to reward subjects and enrich
them. It had nothing to do with invention.  The concept was changed, in
the US constitution, to have a more beneficial purpose, limited to 
inventions, and for limited times.

> (*) Different durations for different kinds of patents. Maybe software
> should be sort enough to make it functionless completely. That would suit me
> fine. I think many good programs behind the scenes do things in non-obvious
> ways, but somehow because someone else stumbles into the same proceedure, it
> just doesn't seem like patentable material to me, at all. 

I think many people would find a 3 year software patent much more
acceptable. But the problem is, once you get into such a short time frame,
it is even harder for the government to keep up. It would be truly
unacceptable to have a 3 year patent granted 18 months into a 36 month
term.  The alternative is then an "accelerated review process". The
problem with that is that it is even more susceptible to bad patents.  
And there is even less reason to challenge them, versus pay them for 3
years.  This also runs afoul of GATT obligations, I think, since the
"normalized" term is 20 years, and there aren't provisions for varying
terms.

Most people think of the patent system in relation to inventors. But the
major users are venture capitalists. They use the patent system to justify
their investments. That is why the patent system can't just be abandoned
or changed capriciously. Doing so would cause economic chaos.  Like any
function of government, the people who have the most money involved tend
to be the movers and shakers of the policy.  If the VC lose interest in
software patents (and there are indications they are), then there is very
little reason to have them at all.  The remaining players are basically
the inventors and the lawyers.  There is an argument that the patent
system protects the inventor from unscrupulous people who simply have
superior manufacturing and distribution capabilities.  This is hard
discount completely, and I think there is a legitimate concern about the
unscrupulous profit on truly novel ideas.  But it clearly becomes an issue
of governance--is it economical for the government or the inventor to
invest the time and expense for a patent application for a product that
has a useful lifetime measured in months?

As soon as you drop the time limit to a major fraction of the processing
time, and lose the interest of the major players, there is little purpose
in having a software patent system.  The value of a true 20 year monopoly
is significant. It justifies significant government and legal expense,
both in creating and overturning, and in defending against infringement.  
The value of a 3 year monopoly is much less so. It basically become a
government jobs program, but there are no computer scientists employed by
the PTO to keep employed. And given the backlog of patent applications in 
all fields, it would be better for other patent system users to not have 
software patents crowding the works.

> XOR operations for a blinking cursor? Can you think of another way NOT to do
> that?

Yes. You can redraw the entire screen.  You can save and restore the
entire contents under the cursor without XOR. You can make the cursor an
analog sprite, drawn over the image as an added analog signal. There are
other ways, but XOR is clearly, _obviously_, the easiest way to do it. It
is not the only way. Its just the obviously best way. There is nothing
novel about the application of existing mathematics to cursors. That is 
expected.

The XOR patent was finally reviewed and revoked.  We used this as an
example of an obviously bad patent for a long time.  But as I said,
mistakes are reasons to change the policies that govern the operation of
the system. They are not reason to change the system itself.  A
knowledgeable computer scientist working in the PTO would never have
approved that patent.

And as I said earlier, it is not the obviously mistaken patents that
concern me most. I am concerned most with the genuinely novel software
patents, and how a monopoly granted 18 to 30 months after its introduction
could impact the industry, the economy, the users, and the programmers.  
A novel software patent is a lot more similar to the airplanes that hit
the WTC, without the murderous intent. The system can be expected to work
perfectly, with disastrous results.

		--Dean