Subject: Re: the walls have ears
From: craig@jcb-sc.com
Date: 1 Jun 1999 14:12:01 -0000

>I just don't like the idea of allowing non-free forks.

I've just taken to calling these "walled" forks, in the sense that code
cannot freely move between them, or back into the libre source base,
and maybe even *programmers* cannot move freely between the development
projects (due to signing various sorts of agreements).

So, I gather that FreeBSD, OpenBSD, and NetBSD are/were forks in the
traditional sense, but not "walled" forks, because anyone who wants
to can copy improvements from one to another (ignoring the usual
technical issues) without having to rewrite from scratch.

But, SunOS and other derivations of BSD'd code are "walled" forks.
While it's true that plenty of innovation can be credited to the
license due to the *aggregate* amount of work going into *all* the
forks thereof, it's also true that those benefits don't accrue
to any one person or organization, precisely *because* so many of
the forks are "walled" (though adding open protocols on top of
some kinds of features mitigates this somewhat, e.g. the cool new
filesystem offered by FooOS can perhaps be accessed via NFS from
the systems running BarOS, though, thanks to them both being walled
forks from the FreeBletch OS, it can't easily be reimplemented in
BarOS.  But the users can't just switch to FooOS, since they already
depend mightily on unique features of *that* system.)

As I said in private email to Johnathan Shapiro, it appears to me that
Linux is getting lots of attention these days, not because it's UNIX
per se (as I have said, the market was *not* demanding UNIX), but
because:

  -  Linux being GPL'ed means that it will not suffer "walled" forks.
     This means people can choose Linux without worrying about
     choosing the "wrong" Linux in a way they never could (and
     probably never will) be able to choose UNIX generally.

  -  Linux being GPL'ed means that it *can* "fork", in the sense that, if
     some entity "takes over" Linux development and goes in the "wrong"
     direction (even if that entity is Linus himself), whoever considers
     that direction "wrong", as *everyone* has the freedom to share
     modifications to it.  (But these will never be "walled" forks when
     GPL'ed code is involved.)

The former is an advantage Linux shares with Microsoft Windows products
(putting aside problems like the Linux/libc/glibc/gcc problem of
backwards compatibility across releases not working, or MS perhaps
intentionally making some kinds of backwards compatibility stop working).

The latter is an advantage Linux shares with BSD/X11 *code* and *some*
products (but not products that are shipped as proprietary, of course;
those are walled forks).

I believe it is ultimately the combination of those two advantages that'll
drive the high-end (informed-user) view that GPL'ed software is always
a "safe" bet, when choosing a platform.  When the risks of choosing the
wrong platform are seen as higher, the GPL will be seen as more safe;
conversely, as in the case of commodity items implementing only public
protocols, when choosing the wrong platform means only a temporary
setback, the GPL isn't likely to be seen as quite so uniquely important.

(Of course, creating products that appear, to users, to be commodities,
but are actually platforms designed to lock them in, is one way to
do business.  MS uses this tactic often, AFAICT.  In the extreme, an
example of this is a company that distributes its branded bottled water
for free, but it contains a virus or other substance that gets the
recipient "hooked" on *other* products from that vendor...products that
aren't free.  I'm quite wary of that effect when it comes to hardware
and software platforms, but, generally, offer me chocolate, and I'll
eat it without fear.)

        tq vm, (burley)

P.S. The importance of branding must not be underestimated, because
the way most of the public knows that the platform they're using
isn't a walled fork is by the name, not the software in it (though
they can, perhaps, be educated to look for a GPL `COPYING' file
that applies to the whole product).

Linux has a bit of branding protection, in that a (perhaps
unenforceable?) trademark for the term is now assigned to Linus
Torvalds (or someone like that).

But only "collective outrage of the netizens" prevents Microsoft
from producing a proprietary product called "GNU" something.

(I find abuse of the ".org" domain to range from the merely annoying,
as in the case of `linuxexpo.org', to the outright unethical, as
in the case of `fortran.org', for the rights to which I *asked* the
business that owned it, only to be told something along the lines
of AlGoreJr's "there's no controlling legal authority".  I don't
think enough net.outrage can be created to retrieve `fortran.org'
for non-profits, and I don't intend to try to do it myself, but,
surely, if MS acquired `gnuemacs.org' and used it to sell a "GNU Emacs",
or something similarly unethical, they'd take a substantial hit to
their already tarnished image.)