Subject: Re: Open letter to those who believe in a right to free software
From: "Stephen J. Turnbull" <>
Date: Fri, 5 Nov 1999 02:01:20 +0900 (JST)

>>>>> "Jean" == Jean Camp <> writes:

    Jean> On the cost of death and open software.  There is certainly
    Jean> a non-zero proability that bad software will cause death, as
    Jean> it has done so in a few cases.

No question about it.

    Jean> There is some probability - increasing over time -- that
    Jean> open source software would be more reliable.

Agreed, but only if you weaken the definition of open source software.
I think it very likely that retaining some proprietary rights in some
kinds of applications can generate enough extra revenue to support
vastly superior reliability (given specified time-to-market) as
compared to free software.  However, some open source practices will
probably improve that reliability (source to users, etc).

I also believe that some applications, eg, OSes, give a substantial
reliability advantage to free software, at least for the mainstream.
(I'm not sure about, say, RTOSes, high security OSes, etc; again given
a strict delivery deadline.  Not that proprietary firms meet the
unrealistic deadlines they currently spout, in general, but best
practice could.)

    Jean> Including this variable could show how sensitive the model
    Jean> is to the cost of life. I suspect not very initially but
    Jean> increasingly so over time assuming software controls
    Jean> increasingly critical systems.

This would be the best way to try to break a empirical model, I agree.
Such sensitivity analysis is important.

    Jean> Further if code is speech,

I think this analogy sufficiently far-fetched that I'm not going to
spend any effort trying to figure out how to model it.  The reason is
that we normally want to prevent authors' speech from being suppressed
in public, while here it is the author's desire to control propagation
of her own speech that is in question.  I don't see offhand how to
create an appropriate "dummy variable" for this.

I think RMS's argument that copyright is domination is much more
plausible, and this does map directly to an easily specified "dummy
variable" as I see it.

    Jean> I suspect you have these results for your model, if so
    Jean> please share them.

I will not have reliability broken out separately any time soon.  That
will have to happen when I do an empirical implementation,
consequential losses are just too big to ignore, but there's no real
point in doing more than a generalized "quality" variable (and perhaps
not even that, depends on how "versions" are modeled, quality may get
disaggregated into non-ordered "features" if I can handle the
horrendous-looking math) at the pure theory stage.

The "RMS constraint" will be in the model at some level.  I might
think about a "code is speech" model, but since I don't understand the 
concept it probably won't fly.

    Jean> There is a nice model of switching cost by Joel West


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."