Subject: Re: street performer protocol
From: "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp>
Date: Wed, 24 May 2000 13:37:05 +0900 (JST)

>>>>> "Crispin" == Crispin Cowan <crispin@wirex.com> writes:

    Crispin> "Stephen J. Turnbull" wrote:
    >> >>>>> "Crispin" == Crispin Cowan <crispin@wirex.com> writes:

    Crispin> Kind of.  I argue that this is the uncommon case (libre
    Crispin> products tend to have multiple vendors).

    >> This is a statistical fallacy (similar to but clearer than most
    >> of your "empirical" arguments).  _By far most market segments
    >> in IT have *no* *zero* *not one* *not even a wannabe* libre
    >> vendor._ Therefore, since all software _can_ be libre, it is
    >> very rare for a product to have multiple libre vendors.

    Crispin> That's a different claim.  I said "libre products tend to
    Crispin> have multiple vendors", not "products tend to have
    Crispin> multiple libre vendors".  I completely agree that the
    Crispin> latter claim is false, which is why I never advanced it.

I see.  I missed that, because I find the claim you want to make
almost entirely uninteresting: it can be just as valid if (as in one
scenario I proposed) the whole libre segment collapses at once.

Eg, because it couldn't fund development and it got crushed by
superior technology.  In today's environment where there is more money
than developers, funding is not a problem.  When the bubble bursts,
that won't be true any more, and we could (and will, I am betting) see
whole segments of libre go from multiple to zero vendors.

By your definition, these conveniently go from "multiple vendors of
libre product" to "not libre product" in one swoop.  No problem, huh?

Nor is the scenario of simultaneous implosion implausible.  Debian,
for religious reasons, I presume, makes it quite difficult to install
KDE.  But I bet they provide a lot of high-quality beta testers for
Gnome, as well as a fair amount of development.  Still, suppose RH's
revenue stream has a (prolonged) hiccup and they decide to cut
Gnome/Gtk funding, emphasize former Cygnus projects, and buy into KDE.
Gnome will survive, but I would bet that this would probably either
force Debian to come to terms with KDE or get marginalized in the
market for integrated desktop OSes.  Either way, the (core) Gnome
vendors disappear all at once.

What I (and I think most FSBers) want to hear about is how OSS can
preserve its presence in markets where it exists today, and extend it
to markets where it isn't yet.  Your claim bears somewhat on the first
issue (although it does not address the countervailing issue of
technical stagnation that should occur due to the decrease in revenue
you point out later in your post---where you ignore, as always, its
impact on future development).  By definition, no bearing at all on
the second issue.

    Crispin> The point I'm pushing is that OSS has an advantage to
    Crispin> customers in that it reduces lock-in.  OSS only has this
    Crispin> effect if there are multiple vendors for the same OSS
    Crispin> product.

I can't believe my eyes.  You mean to tell me that the existence of
Linux, vendor count irrelevant, doesn't reduce lock-in for Windows NT
customers?  The existence of not-quite-OSD-compliant Aladdin
Ghostscript (a monopoly) doesn't reduce lock-in for Adobe Acrobat
customers or other Postscript applications?

(I recognize that this goes in favor of your position that OSS has an
advantage in general.  I believe that---I just don't think your
argument accurately predicts how much, nor where.)

    Crispin> Thus the important claim is that OSS products tend to
    Crispin> have multiple vendors.  That most products have zero
    Crispin> libre vendors is irrelevant.

Not to people who are looking for a market to enter.  Are you
forgetting this is the FS*B* list?  Nor for people who are looking to
escape proprietary lock-in.

In fact, I would argue that these no-OSS marketes are often
_precisely_ the markets where both profitability (to a "monopoly" OSS
vendor) and customer benefits to OSS are highest.  Markets with
existing OSS vendors do not increase options by as much by adding more
vendors as in a market with no incumbent libre vendor.  Furthermore,
adding vendors risks (an understatement) cannibalizing (via price
competition) the revenue stream that funds future development.

    Crispin> My claim, once and for all, for the purpose of this
    Crispin> discussion, is that a single libre product with multiple
    Crispin> vendors has a modest advantage over a single proprietary
    Crispin> product with a large market share (Linux vs. Windows) and

How far does that persist?  Do you mean that given time, on present
form (but assuming moderately intelligent behavior by MS management,
unlike their performance in the trial), Windows will become extinct?
Or do you mean that there are market niches, which you provide no clue
for how to recognize, where that advantage exists and that libre
vendors lucky enough to find them will prosper?

    Crispin> that a single libre product with multiple vendors has a
    Crispin> large (crushing) advantage over a fractured market with
    Crispin> numerous incompatible proprietary vendors (Embedded Linux
    Crispin> vs. {VxWorks, QNX, etc.}.

What makes you think the numerous incompatible proprietary vendors
will stay numerous and incompatible?  It is also true that a group of
proprietary vendors with a common standard (VTR, CD, DVD, cell phone,
HDTV, etc., ad nau.) has a large advantage over the fractured
remainder of numerous incompatible proprietary vendors.  Plus the
revenues to further development within the common standard.  This is
not at all an advantage available exclusively to OSS.

It is true that the OSS bloc has the advantage of core code being
common[1] which _may_ help convergence to standard (RPM certainly has
done that, although Debian persists in its lack of wisdom ;-).  But
this threat may simply cause the proprietary firms to get their act
together soon enough to crush the OSS block with superior proprietary
technology.  So I see the advantage as marginal, even over the next
couple of years.

Finally, I think the idea that OSS vendors sell the "same" product is
overstated.  But this is not the place, except to note that it's
probably easier to incorporate custom, non-published innovations into
embedded products, and there is a strong incentive to do so to
differentiate your product.  (NB. Your customers have no strong reason
to publish; so even GNU GPL products will show sluggish diffusion of
innovations.)

    Crispin> Hmmm ... perhaps that's the theoretical take here: libre
    Crispin> products accelerate the commodization of that product,
    Crispin> and then the market, thus accelerating the tragectory
    Crispin> towards the asymtopic commodity line that maximally
    Crispin> benefits consumers in a mature market.

Or perhaps the theoretical take is that products that are
commoditizing anyway attract libre vendors.  That's the standard
response of proprietary vendors to the Linux phenomenon, as you know.


Footnotes: 
[1]  Yeah, right, in another life I'm <stephen@xemacs.org>: I find
that claim makes my eyes get wet.

-- 
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 straight lines for?  "XEmacs rules."