Re: lynx-dev Re: looking for space for lynx distribution

[David Woolley]

> 
> > >                                                                 and it
> > > should not refer the user to any non-free documentation.  The need for
> > > free documentation is now a major focus of the GNU project; to show
 
> This is the sort of crap which makes me want to run as far away from GNU
> and RMS as possible.

Although I haven't been following the pro and anti GPL arguments very
much, I think you may be overreacting here.  I think the problem that
he is trying to fight here is that a lot of GPLed code is not well
documented, because the authors are programmers rather than writers
as a result a number of authors have started writing documentation,
but have been doing this a as commercial exercise and producing
un-redistributable works.  As a result you have code that is free,
but unuseable.  Moreover, it is probably the coder that invested more
in the work than the documenter, but the documenter is making the profits.

(Interestingly, Microsoft, with all their resources, seem to write such
poor documentation that there is a large third party market - some of this
is people needing documents for pirate copies, but in the office, we buy
third party documentation for properly licensed software.)

Another area he is probably trying to address is de facto bookware, where
the author writes and sells the documentation, but gives software away.

I think one third problem area is that some GPLed stuff, e.g. gtroff,
duplicate large portions of the function of existing AT&T derived
code and as a result, and presumably partly for copyright reasons,
the man page only documents the differences.  To allow people to use
the software without the original documentation, he needs to encourage
people to write complete documents.  (troff is documented in files which
were available from Bell Labs over the internet, but the restructuring
into Lucent Technologies may make these documents disappear.)

I suspect that RMS can't really complain about third party documentation
etc., but he wants the document producers to pay for their own advertising
costs, not, for example, to advertise O'Reilly on every copy of a package.

> According to this `rule', you could not even refer to a commercial Unix
                     ^^^^

It's not part of the GPL, although it may be a requirement for FSF 
accreditation.  I don't think commercial organisations should complain
about this, as they seem to spend more on partnership programs than on
support, sometimes, and often these partnership programs put constraints on
the members - I suspect most of the "best viewed with IE4" stickers are
there to meet the requirments of such programs.

> man page describing that system's own internal idiosyncracies.  RMS
> isn't just trying to provide free software, he's actively trying to kill
> all commercial software entities.  (I work for one, so yes, I'm biased,
> but I don't think I'm unreasonable.)

I've got an email from RMS where he states that he is not trying to
kill commercial software.  I had suggested that the reason SCO didn't
supply certain industry standard FSF tools, except in Skunkware, was
the restriction on bundling, but that got challenged, and eventually I
had a direct exchange in which RMS defined the limits  more precisely,
which actually seemed reasonably generous.  Paraphrasing:  "mere bundling"
extends as far as run time linking, and it is only really static linking
that is a derived work.  You must clearly document the bounds of the
GPLed material and the alternative licencing, but you don't have to ram
it down a turnkey customer's throat.

Given that you have rather higher a profile than myself in this area,
I think it might be worth trying to clarify the position directly.

> As one of those contributors, I'll make the following additional points:
> 
>   *)  If you make this change, I won't be contributing any more code.

I'm not clear wether you are talking about long options flags, or
FSF accreditation.

>   *)  I will think seriously about having any existing code that traces
>       back to me *removed* from the source.

This is a weakness in the GPL; it doesn't seem to give a licence in
perpetuity.  In reality, commercial products can become unsupportable at
the drop of a management hat, but it is probably easier to spread 
fear uncertainty and doubt about the stability of a freeware licence than
it is to point out that a product can be withdrawn in a matter of months,
especially after a takeover.

Incidentally, I tend to see SCO as something more akin to a Linux
distribution creator than an original software generator (I know you
have access to the source and will know the extent to which the freeware
and AT&T original code has been modified, and may be in a position to
challenge that).  Largely SCO is selling the adaptation of the code and
documentation**, to relatively unsophisticated commercial users, rather
than the actual code.

This is probably still too much product oriented for RMS;  I suspect his
model of software economics is one based on selling services to support
and adapt to end user requirements.

There are conflicts between free and commercial software.  I think some
of the reasons people get involved with free software are:

- they can do the job "properly", rather than being constrained by 
  commercial priorities;

- they can produce software that isn't dumbed down or bloated by "easy"
  user interfaces;

- there is not the extreme duplication of effort that results from 
  everyone developing exactly the same thing in secret (I don't think
  this leads to progress, as normally everyone is trying to do the same
  thing, rather than their being a continual leap frogging process).

As markets saturate, I think commercial software can go either in the
direction of increasing gimmickry with little useful added value 
(DOS Word 5 runs faster on an 8086, and is more stable, than Word 97 on a
Pentium 90, and still provides more functions than most people ever use),
or the move has to be towards building customised solutions from standard
software components, and supporting them.  The catch is that software
components have a design cost, but virtually no manufacturing cost,
so, once you have developed them once, ones that will do the job become
essentially free, unlike nuts, bolts, gears, resistors, SSI logic gates, op
amps, etc., where there is still a real manufacturing and materials cost.

From time to time, people try to indstrialise software development, to make
it easier to manage, but in either of the above scenarios, the cost ends up
in the non-indstrialisable parts of the design process.

** In the past I've had to use SunOS documentation to understand SCO UUCP
and the documentation in the original source package to fully understand
their MMDF.  This is essentially because the documentation has been 
editted down for the target market.