Subject: Re: street performer protocol
From: "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp>
Date: Wed, 17 May 2000 12:28:07 +0900 (JST)

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

    Crispin> "Stephen J. Turnbull" wrote:

    >> Sorry, I should be more clear.  I don't consider a Linux
    >> distribution a product of software development.  The product is
    >> (nearly) all packaging.

    Crispin> But this too defies empirical data.  Many (if not most)
    Crispin> Linux distributions do some kind of software development
    Crispin> so that they can have some kind of value-add to
    Crispin> differentiate themselves.  Some add proprietary code
    Crispin> (Corel Word Perfect) while others add GPL'd code (Red
    Crispin> Hat's RPM and GNOME).

Closed code is not data of interest here.  Proprietary public code may
be (I certainly would count it, but at a discounted rate because of
the implied friction to user-contributed improvements).

Looking at the history[1] of the semiconducter and biotech industries, I
would predict that Red Hat's specialization in GPL'd code will soon
end up in the garbage can.  (Presumably it already has, unless they've
GPL'd Code Fusion.)  How far they go down that slippery slope is
another issue; maybe they can even climb back up. :-)

    Crispin> This is the open source re-use effect at work: because
    Crispin> each of these 100 companies does *not* have to re-invent
    Crispin> the wheel, they each are free to work on some special
    Crispin> feature that is not currently addressed by the community.

And the ones the prosper will be the ones that add it in a proprietary
fashion, at least at first.  That's my argument in a nutshell.[2]

    >> Amazon.com knows nothing about publishing.  Linux distributors
    >> need to know more than that about development, of course, but
    >> their core skills are (IMO) closer to Amazon's than to Sun's.
    >> Only a few will survive as pure free software businesses.

    Crispin> I don't believe that many companies will qualify as a
    Crispin> pure free software business by your terminology.  AbiWord
    Crispin> is a notable exception, but it's not clear that they will
    Crispin> succeed.  It is much clearer that a hybrid model (part
    Crispin> packaging, part development, all free software) can
    Crispin> succeed.

Sorry again.  It's that last strategy I meant when I said "few will
survive as pure FSBs."  Ie, I don't think it is any clearer than a
typical Alan Greenspan pronunciamento.

Unless you consider the purely financial rearrangement that an IPO
represents "success."

At the moment, producing a Linux distribution is hard work all by
itself.  You need to fix bugs, then back out your own fix at the last
moment when the upstream projects deliver canonical fixes.  You need
to arrange for coexistence of products that compete for similar
resources, both having fanatical adherents (GNU Emacs and XEmacs, or
any subset of window managers you want to mention).  But this is going
to get easier (or the LSB is meaningless); there will be cost
competition.  The low cost winners will have more funds to do
specialized development, get bigger, then start banging heads with
each other (we see this in "cluster solutions" already).

    Crispin> Or a strategy focussed on packaging, distribution, and
    Crispin> service.

    >> Explicitly abandoning development, I see.  That's not what we
    >> want!  We want people making _big wins_ for _creating_ and
    >> _developing_ products and contributing their source code to the
    >> community.

    Crispin> That's what a hybrid model would seem to deliver.

All I see is (a) an absurdly overvalued stock market providing (b)
absurd amounts of funds to a project (OSS) we would all dearly love to
see (c) succeed.  (a) is in the process of rectification, apparently,
which makes one suspect that (b) will soon be history, casting a dark
shadow over (c).

In such a rapidly changing market, I think your facile "empirical
facts" are dangerous to rely on.  I don't see fundamentals for growing
open source development anywhere in current practice of Linux
distributions; I see firms scrambling after niche after niche.

Capital is flowing to firms like Red Hat in part because of their
visibility, ie, portal-like role.  I don't know that this is a stable
foundation for a Linux distribution anymore than it is for a Web
browser maker.  Where do you see the fundamentals to grow strong open
source development organizations?

I think the best hope is organizations like Cygnus, with their roots
firmly in several complementary fields: shrink-wrap proprietary
solutions ("cash cow"), high-value-added free software consulting (ie,
"Jonathan's market"), and infrastructure building (mostly pro bono,
but also constitutes an investment in the consulting business).

Proprietors with Brian Bartholomew's tastes can of course eschew the
shrink-wrap market (or adopt the "3M strategy" with open source
shrink-wrap products:  milk it while it's new, then "abandon" it ==
support but don't waste good money by further developing it---I don't
think he'd much approve of that strategy, either).


Footnotes:
[1]  The analogy is financial, not technological or in terms of IP
structures.

[2]  Not a Nutshell[tm].

-- 
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."