Subject: Re: Why software patents are bad
From: "Stephen J. Turnbull" <>
Date: Thu, 23 Sep 1999 16:26:51 +0900 (JST)

>>>>> "Bernard" == Bernard Lang <> writes:

On Wed, Sep 22, 1999 at 05:31:55PM +0900, Stephen J. Turnbull wrote:

    >> It is _exactly_ how it is accomplished that matters.  I really
    >> can't imagine that French law differs that much from U.S. law
    >> in this respect.  You can't patent an idea until you can
    >> demonstrate how to implement it; it is that demonstration that
    >> justifies the award of a patent, and only for the
    >> implementation, not for the specification.

    Bernard> I thought so. But I made the argument that patenting
    Bernard> encourages hiding code (so as to avoid infringement
    Bernard> detection) and that infringement actions are more likely
    Bernard> to concern interfaces and protocoles. His reply:

Well, if you can patent a protocol ... but that would be like
assigning IP in look and feel, which is not permitted in the U.S.  Of
course this is a very fuzzy area, since protocols are algorithmic at
both ends, whereas look and feel is not.  (AFAIK)

And a claim of infringement on an algorithm (eg, RSA) would require
looking at the insides.  Ditto "XOR cursors" (or maybe not, I dunno if
mathematical tautologies are considered equivalent, since the
operations are different).  Diffie-Hellman (I'm not clear on whether
they actually got a patent; assume they did), on the other hand, would
be upheld if you were caught distributing a public-key system, unless
you could show it somehow didn't fit their framework.

    Bernard>     The availability of source code has never been a big
    Bernard> factor, at least here in the US.  There is a governing
    Bernard> principle, Doctrine of Equivalents, which says basically
    Bernard> that if you are doing the same thing in a similar way,
    Bernard> you still are infringing.  So if I have a firewall
    Bernard> application checking on incoming addresses, it probably
    Bernard> won't matter how I do the checking, if I am claiming
    Bernard> firewall protection in general.  For this case, access to
    Bernard> the source code would be ireelevant.

Claiming firewall protection in general probably won't work, it's too
broad and there would probably be prior art.  You could claim a
firewall based on checking addresses; then it would not matter how the
address was checked ... probably.  Except that ident (RFC 914, isn't
it?) would be likely prior art for checking sources in general.  RFCs
821 and 1123 would be prior art with respect to flitering on the
address in HELO (the fact that RFC 1123 prohibits it notwithstanding,
AFAIK).  So you then have to be very specific that you were filtering
on the IP address in the IP packet header.  And now source code is
indeed irrelevant, because it's trivial.
If you want to argue that the Patent Office does a terrible job of
checking the prior art for software (and thus might issue a patent on
"firewalls in general" or "address-filtering firewalls"), I can't
disagree; I haven't the detailed data and the anecdotes are pretty
horrifying.  But that is quite different from saying the system is
fundamentally flawed, at least in terms of gathering the necessary
political support for abolishing software patents.

University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
What are those two straight lines for?  "Free software rules."