GPL not encouraging new technology

[Niall Douglas]

On 1 Dec 2002 at 21:44, Marcus Brinkmann wrote:

> > > So there is your misunderstanding.  The innovation in the Hurd
> > > lies in the fact that users can do this extension, without asking
> > > Microsoft, or even the system administrator, and without
> > > compromising the security of the rest of the system. 
> > 
> > But that's not innovation - that's merely a consequence of the Hurd
> > being "free" software. On a technical level, the two are capable of
> > similar functionality.
> 
> Another misunderstanding.  "user" above means unprivileged user of the
> system, ie a user that has not the user ID 0.

Ah I thought you meant user = human, not program.

> See, I am not on a mission here.  If it's important for you to believe
> that no innovation is possible in free software, you are entitled to
> your opinion and I am not interested in challenging them (I will be
> interested to see your upcoming proposal).  I don't even care if you
> consider the Hurd to be innovative, ex-innovative or not innovative at
> all.  But, and I insist on so much, on a technical level, it's
> different from Windows NT. That all operating systems are capable of
> similar functionality is a trivia (you only need to define "technical
> level", "capability", "similar" and "functionality" conveniently
> enough).

Ok, returning to what I originally said, which is that it is somewhat 
similar. Of course similarity is subjective, so what I might find 
similar is not what you find similar but then again, you're so very 
much closer to the code. I may not understand the GNU Hurd technical 
descriptions and indeed it's entirely possible the NT kernel 
technical descriptions are just plain wrong, but I personally doubt 
both. Of course there are a number of things GNU Hurd does 
differently than NT, but if I were to compare the current Linux 
kernel against GNU Hurd and the NT kernel, I personally would say GNU 
Hurd is more like NT than Linux.

And just for the record, I *do* think the GNU Hurd is innovative, and 
indeed I have said just that here a number of times. I am very much 
looking forward to seeing & using it when it becomes the de facto 
kernel.

> If you want to know why the Hurd didn't happen earlier, you also have
> to look at the history of operating system development.  Mach was only
> the first generation of a microkernel, and it's far from perfect. 
> That led to a situation where the whole world turned away from
> microkernels and focussed on monolithic kernel approaches, or at least
> single server systems.  This meant that microkernel research went into
> a depression, as well as its application in industry.  In fact, some
> people claimed that operating system research is dead
> (http://www.cs.bell-labs.com/cm/cs/who/rob/utah2000.ps). Now that the
> hype is over, hardware is a lot faster and cheaper, and features that
> are naturally included in a well-designed multi-server operating
> system become important for users, we have the chance to take another
> look and "do it right".  It's really a very classical hype curve, and
> has nothing to do with free software at all.

That is interesting, but I think that history is mostly relevent to 
the Unix world. I'm not going to argue the point however, because 
quite simply I don't know much about that period of time. If you say 
that was the way it was, then I believe you.

Alex Hudson wrote:
> > > I don't think that's necessarily the case. Proprietary software
> > > tends to come to market fully-formed (i.e., developed already);
> > > Free Software isn't usually written and then marketed.
> > 
> > Now I'll repeat myself, this time with hopefully more clarity.
> > Proprietary software tends to get money much earlier in its
> > development cycle because the workers need paying. Free software
> > tends to get money only once it's mostly complete and mature because
> > you can't sell something which isn't written yet.
> 
> Here again you've missed what I said. I disagree with that fundamental
> assumption you make above. I'm not going to argue with the rest of
> what you say; there's not point, because I believe your assumptions
> are wrong. You state the above two points as if they were facts;
> they're not. Unless you are stating your opinion? 

Right, we're finally making progress! Why didn't you just say my 
assumptions are wrong?

Of course, anything not backed up by references and academic 
statistics is going to be opinion and not fact, and indeed I'd say 
even with references and statistics much will remain opinion. I had 
thought the notion that free software attracts less money than 
proprietary to be self-evident, but if you don't think so then I 
suggest you look at all software created world-wide and look at where 
the money goes. Why isn't RedHat in the position of Microsoft? Or are 
we claiming that in not much more time, it will be?

Copious historical evidence would show that if you have money, you 
can buy the best and the brightest of any speciality and get them to 
do what you want. That means rich people tend to stay rich and indeed 
get richer. Now if right now I had a few million in the bank and I 
decided I'd like to design the perfect operating system, I could just 
employ the best and make it happen. If I'm poor without a penny, then 
the only people I can attract are volunteers, most of whom will have 
their own wildly different ideas and so won't work with me.

People like working with what they know and best of all it's common 
ground. For the opinions of the best future of any engineering, 
you'll get wildly different options and traditionally it was the role 
of management to choose one. Free software doesn't have management, 
it has the democratic process - if two groups disagree, they'll fork 
off their own version of the software and competition will determine 
the winner. This is all good, but not for radical blue-sky ideas 
because the further you get from the standard, the more widely 
differing the opinions of the optimal become, and the more impossible 
for a cohesive group to form.

I think my logic's correct, but I've used more assumptions. Please 
tell me those you think wrong. It's basically the too many cooks 
spoil the broth argument.

> > This is all true and I completely agree the patent system is not
> > perfect. However if you consider the world before patents, the
> > overall situation was worse.
> 
> You're saying that the patent system was responsible for the end of
> the dark ages? I think most historians would argue literacy was, but..
>  I don't see how patents were a cure for anything. They were, and are,
> purely a device for making money, the 'fosters innovation' argument is
> tenuous.

Umm, never said it ended the dark ages and I would point to italian 
politics and protestantism as being the key to the end of the dark 
ages, not literacy (which other cultures had and have, but it doesn't 
help them).

Of course patents are there to make money, but more specifically to 
create an entrepreneurial spirit. The average guy on the street 
probably invents something new every week, but if there's no 
incentive to spread their ideas then they won't spread. Now I'm the 
first to say the current IP system is crap, but before you throw the 
baby out with the bathwater you have to look at what happened before 
they were introduced: big companies stole ideas from startups and 
could drive competition into the ground by producing a clone with 
much lower overhead. This discourages the startup, thus never 
challenging the big company and thus creating a monopoly.

The precise reason the IBM's and Microsoft's of this world can exist 
is because these checks and balances are broken for software. What is 
the case for a physical device (materials, manufacturing time, 
processes etc.) is completely different to software (almost nil cost 
for replication) and the original mistake made was to make software 
equal IP design whereas in fact it's IP implementation.

> I suspect we have different experience (you mentioned the military).
> In my experience, especially in modern times, companies are very
> unwilling to give up 'IP', especially if it's something they can go on
> to develop further.

Yeah the military gets these notions of them being unable to get bugs 
fixed during war. Such a thought always makes me smile, because if 
they wouldn't use COTS they'd be helping themselves a lot there. I 
actually proposed once that we make the entire solution on FreeBSD 
and literally they looked scared :)

> > That is entirely possible, and of course I know the proprietary
> > range much better than the free one. However I came at this point
> > from the basis of logical theory and then looked for evidence to
> > disprove it. So far, I have not found any.
> 
> But, what I'm saying is that I don't think you have set out any case
> by which you can disprove it. For a start, trying to disprove
> something is exactly the wrong way to go about things...

Well that's what the logical theory books also say, but I'd wonder 
how many people really ever start with a tabula rasa?

> In your responses to my examples, the first was a question (e.g.,
> 'maybe' :), one was an example demonstrating you don't realise what's
> 'wow' about the Hurd, and the third was arguably a straw man (rather
> than argue that Apache 2.0 multiprotocol support isn't innovative, you
> argued that developers are inherently conservative and hence a project
> couldn't be innovative). I can't read your examples any other way.
> Personally, I believe this to be bias on your part. Of course, since
> it is an unwinnable argument, neither of us is correct.

No I think it's partially that I'm not explaining myself. Basically, 
I'm working from a belief that a complete ground-up redesign and 
total rethink from basic principles every five years is a very good 
thing for software.

Now I don't see much of this in the free software community, at least 
not much with a lot of volunteer energy behind it. I do see more in 
the proprietary world, because they're making long-term bets.

I know Microsoft, AT&T and IBM all have done ground-up rethinks, many 
of them. Whatever we all may have against them for whatever reason, 
surely we can agree more radical blue-sky (ie; ground-up 
reconceptions) work is done in the proprietary world?

> > It has everything to do with it. Government subsidies are AFAICS the
> > only method free software has of creating real innovation.
> 
> Why is business subsidy impossible? If we're arguing about subsidy,
> subsidy is a method that pays for Free Software before it can come to
> market - precisely how you compain Free Software is funded :/

Because business or government subsidies of this form aren't 
entrepreneurial - they're effectively service payments. While they 
could be entrepreneurial, it isn't likely because once released, 
someone else can run away with their investment.

> > DARPA's funding of ReiserFS. There is next to no chance of pure
> > blue- sky research in free software, whereas if nothing else
> > Microsoft's many failed blue-sky projects show some effort is put
> > that way with the hope of first-to-market.
> 
> Nautilus is a great example of a blue-sky project that failed. And
> unless you used Eazel services, please don't tell me it wasn't
> innovative :p

I looked around, but couldn't find much about Nautilus except that it 
was a file manager for GNOME. I note with interest it failed to 
secure additional funds and so failed, which reinforces my point I 
think (though I'm sure you'll say KDE was more to blame).

> I personally believe a lot of the Hurd is *very* blue
> sky - in the vm example I gave you, I think it's highly innovative for
> processes to manage their own memory needs without requiring ring 0
> CPU access. And this isn't old; this is new Hurd innovation. 

For the umpteenth time, I *do* think the Hurd is blue sky and 
innovative, just not as much as it was five or ten years ago. I keep 
repeating this, but no one's understanding. Go back and check the 
emails, I *have* said this.

> > Oh come on now! You telling me free software isn't on average of
> > much higher quality than proprietary? I'm sorry, I just don't
> > believe that.
> 
> No, that's not what I'm saying, although I'm not necessarily agreeing
> with you. I'm saying that Free Software isn't necessarily of high
> quality. There's a difference there. If something is Free Software, it
> doesn't gain peer review because it's free. It gains peer review
> because people are interested in it.

*And* because people have access to its source.

> Free Software is not a
> development model, and hence cannot affect the quality of software.

You're wrong here, free software strongly influences a development 
model and IMHO, that influence produces better quality results.

> Free Software tends to be more pervasive, yes, but I'm not convinced
> that software + more developers = better software (mythical man month,
> and all that jazz..)

The calibre of a programmer wanting to volunteer their time for no 
fiscal reward tends to be much higher. If you've seen some of the 
people coming out with degrees nowadays, you'll know the average 
calibre programmer is pretty low but that's basically because of lot 
of them take the degree not because they like the subject, but 
because it has guaranteed jobs.

> Hurd not attracting programmers? I think it's to do with a number of
> factors, but you'd be 100 times better asking this question of Marcus,
> or anyone else with more than 10 minutes use of the Hurd (like me ;).
> I think one reason is that there aren't many hackers with enough
> low-level hardware knowledge/PC boot knowledge to go around.

That's understandable, it's nasty.

> The main
> reason is probably that it's a system people don't have much knowledge
> of - Linus had Andy Tanenbaum's book, for example, new hurd developers
> have much less documentation to go on, and obviously less relevant
> experience because it's such an innovative project. I also think that
> the FSF/GNU puts off a lot of people, because they don't understand
> the organisation.

These are all points I made ...

> Ah, so you are holding Free Software to specific standards that a
> "commercial" (proprietary) venture isn't held to. You state you think
> Free Software is always commercialised as a mature project (e.g.,
> subsidised), you also state "every commercial venture will subsidise
> any new programming project", but then you say to your standards free
> software innovation has to be unsubsidised for it to count. If that's
> not what you mean, I apologise, but I'm sorry, that's how it reads.

That is what I mean. I am holding them to different standards because 
they are two totally different beasts. One is all about primarily 
generating profit whereas the other is primarily about generating 
software. It's in the nature of proprietary to subsidise new works 
whereas it's not in the nature of "free" software - there it's a case 
of enough enthusiasm pointing the same direction.

> > Quite simply, I'm asking the question: who will produce the
> > most radical innovation over twenty years? Free software or
> > proprietary?
> I would go back to my point on competition: if innovation were
> possible, and it was desirable (from a consumer's point of view), and
> only proprietary software was capable of delivering sustained
> innovation, then proprietary software would be able to out-compete
> Free Software because it would be able to out-innovate it. 

You're not quite right there: proprietary will only innovate when 
it's more profitable to do so. Free software will trickle-innovate as 
part of its nature.

I think it's a question of two points: do you believe complete 
rethinks are good for software? If so, how many naturally occur in 
either proprietary or free?

I say more happens in proprietary.

> Well, clearly you are arguing for something non-free, although I doubt
> you dispute that. I personally don't see a workable legal basis for
> what you describe, even if it were desirable: the only means I can
> think of make it equivilent to shared-source, which is not better than
> what we have. I would like to see what it is you propse, merely out of
> interest :)

The link should be posted shortly.

> What you describe just sounds like non-free software, to me. And it
> will suffer an inability to complete, just like current non-free
> software. 

Oh you're right, it is non-free. But it's also free. You'll see.

MJ Ray wrote:
> > No, but money arrives to GPL projects much later (and less of it) 
> > than to commercial so therefore proving my point.
> 
> [etc etc ad inf... the above is just a typical quote]
> 
> Please, stop waving your hands and bring up numbers.  If you have, I
> apologise but I missed them in the volume of email, so could you point
> me towards them?

No I have no numbers whatsoever! And I do apologise if I'm boring 
people. Hopefully above I've explained why I'm making the assumptions 
I have.
 
> Also, you still seem to be stuck in the conventional software models
> despite accusing others of wearing blinkers.  Please help me resolve
> the apparent contradictions in your reply.  Concise clear replies to
> each thread individually in turn might make your writing less
> confusing.

Well I wasn't wanting to be filling the list with my replies, so I 
was keeping them all in the one reply. I fully understand the free 
software business model (it's a service), it's just (a) I don't 
believe it makes much money (b) it isn't good long-term for software.

I know many in here will prefer the model of software writing being a 
service. I disagree and think it should remain manufacturing - if 
anything, Maggie Thatcher showed us why the service industry isn't so 
hot.

> Of course, you may not care whether anyone else understands your
> thoughts. If that is the case, please disregard my grumble.

Well like most people, I'd prefer to think at least someone 
understands me :)

> > Proprietary software tends to get money much earlier in its 
> > development cycle because the workers need paying.
> 
> This does not change for free software if you are developing it
> commercially.

I think it does. Capital goes where it can make the most profit, and 
free software does not make much profit therefore it's a bad 
investment.

If you make software construction a service industry, then money must 
regularly come in because it becomes a running cost. However this is 
bad long-term, because then we directly compete against the third 
world. This is a very very bad idea because they'll trounce us - they 
can do the same work, but cheaper.

The best long-term option is to maintain ourselves as an innovation-
based entrepreneurial manufacturing industry.

Simo Sorce wrote:
> > Now I'll repeat myself, this time with hopefully more clarity.
> > Proprietary software tends to get money much earlier in its
> > development cycle because the workers need paying. Free software
> > tends to get money only once it's mostly complete and mature because
> > you can't sell something which isn't written yet.
> 
> There's no proof of this theory.
> It's your idea.

I think economic theory is on my side here. We may not like 
capitalism, but it is the system.

> > The only source of 
> > substantial funding for infant free software is government grants,
> > and those really are bespoke.
> 
> It is the same for "real" blue-sky wannabe proprietary software.
> I know no company that invest money on a project that need many years
> of development. They all depend on previous governmental university
> research.

That's strange, because I know of many companies who have invested 
years in a completely new product designed to take over after their 
current one.

I do agree most blue-sky research is done in collaboration with a 
university so that a company may get the government to pay for some 
of the work, and I personally think it's the worse for it.

> Tell me of a real "blue-sky" project that was completely developed and
> funded by a company and that involved a substantial amount of
> innovation that made it something completely different from what there
> were before.

Ok, there's one that I know well, which is ARX from Acorn. There's no 
doubt it was mold-breaking in its day, and indeed perhaps it was 
overly so since it ran out of cash and they had to go make RISC-OS in 
six months. Nevertheless, RISC-OS reused much of the innovative stuff 
written for ARX eg; font manager, filing systems. However, had ARX 
survived, it had a memory-protected preemptive multitasker and the 
entire system was written in a form of object Pascal. It should have 
been superior to Amiga's WorkBench which did some pretty innovative 
things on a home computer.

I would also say DEC VMS was pretty innovative, and the NT kernel 
does things nothing else in production use before it did. Lastly, and 
I know people will disagree, but what Apple has done with MacOS X is 
precisely the kind of venture project proprietary software can 
produce - when a company's looking at death, it can be motivated to 
spend on some really radical ideas. There isn't an analogy in free 
software, because a developer can't really die like a company can 
(not with the scale of loss to the board of executives anyway).

> I think you will not find out much, as proprietary software is tied
> much more to wide public acceptance and so imitate much more what
> already exist.

I'd strongly disagree with that. There isn't the same raw survival 
instinct or entrepreneurism in free software.

> The real big innovation of the last 2 decades was the net, and it was
> not made by companies with a proprietary model, or it would have gone
> nowhere (see IPX, DECnet, SNA ...).

Object orientation? Virtual Machines? Wireless communications? The 
list goes on and on. Conversely, the internet is more or less the 
same as DARPA had it (which precisely why it has technical problems). 
The proprietary alternatives couldn't win because the owners had 
vested interests in their technology which prevented their wide 
acceptance eg; like with minidiscs.

> > And that's my problem with free software - it is not long-term self-
> > sustainable because it cannot encourage entrepreneurs except those
> > making money off the back of selling other people's work. This is
> > bad (in the long-term) for software and hence society.
> 
> Unproven, and misleading, imho. Why selling other people work's should
> affect software production? I see no connection, is like saying that
> if bakers didn't existed bread will have never be done.

I never said it affected production. I said it affected innovation. I 
also said it discourages entrepreneurial investment.

> Software is a service, you do not need to sell it as a box to make a
> living, there are many ways people live with free software providing
> services.

No it bloody well is not - it's a manufactured product. It's 
precisely that kind of dangerous thinking which is taking us to a 
very dark world indeed where every algorithm is patented and the 
third world writes all the software while we're all out of jobs.

Just because a product can be replicated for near zero cost doesn't 
remove its high design and first copy implementation cost. Software 
is a manufactured product. Software *support* is part service part 
manufacturing like say a plumber is.

It is *not* a service and it's very very dangerous to think of it 
that way.

> I think you go wow because you know not much free software and tend to
> think (as many): "oh dear how do unpaid(?) people manage to produce a
> so useful software?" While for a proprietary software you give as
> datum that an overpaid software must have quality and you look only at
> features.

I would take exception to you saying that about me. I have offered no 
evidence to support this conclusion of yours and in fact I find it 
offensive. You are merely labelling me with your preconceptions 
because I don't 100% agree with what you think is your dogma.

I have always judged software purely on its technical merits without 
regard to cost, who wrote it, if it came from a terrorist country or 
whatever.

Furthermore I think you'll find I have released a great deal of my 
own software for free over the years. And I don't mean GPL, I mean 
completely free.

> > Daemons no but plugins yes. The NT kernel is quite extensible eg; it
> > uses a unified namespace of which parts are provided by plugins eg;
> > pipe manager, file system etc. Technically one could do much of what
> > GNU Hurd can in NT - it's just Microsoft have chosen not to and
> > indeed seem to actively have prevented anyone knowing much about it.
> 
> No, no, no, please can you try to document yourself before making
> assumptions?

Very hard to do regarding NT as so much of it is undocumented.

> The Hurd is completely different, they are indeed based on a
> microkernel architecture, but here where the similarities end.
> 
> We speak of daemons not of plugins for a real valid difference:
> in hurd you can plug in a new filesystem as user of the system, you do
> not need to be root to do that, and this is something NT kernel, nor
> any other classic kernel can do.

So the filing system executes in user space rather than kernel mode? 
I don't see the ground-shattering significance.

My original point in all this was about *similarity* ie; how 
similarly structured are the two? I said somewhat similar and so far 
everyone here has said completely different without saying why. So 
really, you're all just as guilty of not backing up points as much as 
I am!

> If you honestly answer these questions you will see that proprietary
> software is conservative as well, and that there has not been much
> innovation in this industry besides the success of Free Software.

I agree, but then all fields of human endeavour tend to be 
conservative. The difference is that proprietary spends a small sum 
every year on the really radical stuff as a long-term bet and I can't 
see an equivalent in free. Also proprietary companies can become very 
radical when they are starting at bankruptcy and I again can't see an 
equivalent in free. Lastly, proprietary has some entrepreneurism 
whereas free has very little.

> Well, there's always need
> for new software, everything need software nowadays and any new
> product or process need new software or need to adapt existing one,
> I'm sure developers will have a job in future in a free software
> world.

Capitalism works squarely around growth. If you're growing, all is 
well. If not, all is bad. That is capitalism. What you are proposing 
is we reduce the drive for growth within our industry which in the 
long-term means less jobs, lower income, higher profits for the 
multinationals and less innovation.

I personally think capitalism is a doomed system, but I at least know 
its first principles. There are some excellent books on capitalism 
eg; R.H.Tawney's Religion and the Rise of Capitalism along with why 
it's doomed by Fritjof Capra.

> No, "on average" it is not better than proprietary.
> Quality is not something inherent to free software, a license cannot
> change the quality of the software by magic. Go and look at
> sourceforge and see. You will find very good software and very poor
> software, I'm not sure about the average.

I'm sorry, in my experience this is not true - free software is of 
higher quality, just the same way as RISC-OS software was generally 
of a very high quality due to its homebrewed nature. Philosophies 
directly affect quality, not just in software but in all areas of 
life eg; fascist governments tend to increase quality, communism 
tends to decrease it.

> look at the last 20 years of free software and think that now people
> producing free software has increased at exponential rates.. go
> figure!

Only because the current proprietary model is so crap. A different 
one would 
draw a lot more people back to it because people are selfish.

> >  and (c) support from politicians in order to bring it 
> > about.
> 
> We are seeing support here in Europe

Really? Then why are software patents in Europe being even talked 
about? 
Why are there fast track visas for software programmers? Why is there 
a singular resistence to broadband? What about the RIP acts being 
instigated all over Europe?

Sorry, I don't agree. We're just as screwed as anyone else without 
donation funds.

Cheers,
Niall