Re: [HACKERS] BSD vs. GPL

At 01:15 20/06/99 -0400, Bruce Momjian wrote:
>
>I have followed this discussion, and while there is a lot of it, there
>aren't many specific examples.  Let me suggest one:  BSDI
>
>They took the BSD4.4(386/BSD) code, hired many of the departing BSD
>folks, and developed BSD/OS based on it.  Now, if the BSD code was GPL,
>would those people have started a company...
>
...
>...  Only "Source licenses holders" receive
>OS/kernel source, which currently costs an additional $2k.)

This is one model for development. There have already been counterexamples to it being the only model. My personal
experienceas a consultant and developer suggests your model is probably not even the best. I have had occasion to use
thesupport services of a number of the larger IT companies, and while the support can be very professional, it leaves a
lotto be desired:
 

1. No commercial company will say "The optimizer code is stuffed, we're not sure how to fix it but we're working on it.
Usea work-around". Instead they will leave you in the dark, saying "It has been passed on to engineering, and we'll let
youknow when it is fixed" [the last part only sometimes]. For me it is far more useful to know that the bug is a real
problemthat may take a while to fix. Open source even gives me the *choice* to employ someone to fix the code
(especiallyfor those products that have commercial support), and I am happy that such fixes be made public.
 

2. Access to the developers of commercial software is rare, and controlled even at the best of times. When you do get
totalk to them at a conference, for example, they are totally unwilling to talk about futures. Contrast this to Linux
orPgSQL. Future plans for a database are very important, and affect my choices in database design.
 

3. Quality. There are many aspects to quality; the most important in my view are: does it work? when it doesn't work,
isit easily fixed? OK, PgSQL is not as reliable or robust as Dec/Rdb, but I am not a 24-Hour shop with mission-critical
applicationsrunning all the time, so I don't need Rdb. Experience with bugs in PgSQL and in other database products
suggeststhat PgSQL bugs get fixed quicker. Frequently a new patch appears within a day of a bug report. Patches for
knownbugs can be downloaded from the mailing list immediately. Linux is another example; it is more reliable than NT,
andagain, the few bugs that I have reported have been fixed in days. Contrast this to Microsoft's support for NT! For
theGPL products I use, I would say the oevrall quality is higher than commercial offerings.
 


>Consider Redhad, Caldera, etc.  They are adding value "on top of" the
>OS, but the kernel is pretty much the same for all of them.  In fact,
>aside from some tweaks, they really aren't involved in enhancing the
>lower levels of Linux, and economically, they really can't.  They could
>put 100 programmers on it, but once they do a release, all their
>competitors have all their enhancements, and the economic benefit of
>those 100 programmers is gone.

Others have already pointed out that your facts are wrong here, and so too is the philosophical point:

For an existing *large* GPL project, any additional code developed by 100 programmers will require 100 programmers to
maintainand enhance. If the original programmers are all fired, then the product becomes unsupportable and worthless -
youwould be very unwise to buy it. 
 

Furthermore, since most of the money for GPL'd products comes from support and ancillary sales (eg. commercial products
basedon the s/w in question), for anyone to become competitive with the original developer they would require a
substantialinvestment up front (to understand the code), and a continuing investment in development and support. 
 

Rather than compete they would be better off enhancing some other part of the software and thereby developing their own
niche.All parties can then use the improved software. This approach makes the product stronger, so increaes market
share,and solidifies the basis of the two companies.
 

>Sure, Linux is better for it, but those
>100 programmers aren't seeing an increased sales rate to pay their
>salaries.

This is true; many 'volunteer' programmers do not see fair monetary recompense in the short term. But the small amounts
ofunpaid work I have done has been a good learning experiance, enjoyable, interesting, and made me feel good about the
workI do. This is at least *some* compensation, ignoring for the moment the competitive edge such experience gives me
inthe market place. Combine this with the fact that many commercial companies see the value in paying their employees
towork on GPL code, and I think you will find that 'those 100 programmers *are* seeing an increase *in their
remuneration*.In some cases they may only have jobs because of it.
 

>So, the GPL vs. BSDL issue really boils down to whether a particular
>piece of software is going to need a commercial organization to
>improve/enhance it in the future.  

No, it does not. It boils down to whether or not the internet community is large enough to continue to produce high
qualityvoluntary contributions to projects. There are always people and companies who are unwilling to work for
anythingother than hard cash, especially in the current economic millieu, but GPL will work as long as developers see
valuein GPL.
 

I am lucky: I get paid reasonably well for the work I do, and have a lot of work (at least, at the moment!). This means
Ihave the luxury to be able to contribute my time (in small ways) to selected voluntary 'clients'. These are generally
organizationsthat have little or no money to spare, and could not afford a programmer under any circumstances. I could
nothelp them if products like PgSQL and Linux were not available and of such high quality. And I strongly believe that
theywould not be of such high quality if the source was not open.
 

>If it does require a
>commercial team that can put man-years into the project and needs to
>recover the costs of doing that, GPL will prevent that from happening,
>and a commercial entity will have to start from scratch in developing
>the code so they can "own the code."

This is clearly not true, as has been argued elsewhere.


>The answer to that question also suggests the question of whether
>non-paid developers are the future for "all" software.  

Even I don't suggest it's the way for all software. Our current social and legal structures pretty much require that
someonetake responsibility and due care for some things (eg. heart monitoring software). The best way to show 'due
care'is to buy commercial software from a reputable company. The *best* software may still be freely available on the
internet,but the safest software (ie. readily available for legal action), will always be commercial.
 

>If it is, then
>GPL is the way to go.  If it is not, then GPL use needs to be decided
>carefully depending on the perceived need for later commercialization of
>the code.

No, it depends on the perceived niche for the code; if I come up with a hardware/software-neural-database-thingy, then
I'mNOT going to make the software open source - it would disclose commercial secrets. Similarly, if I am the sole
developerof a complete, high quality, working product, then I am *inclined* to keep it commercial - but ONLY if I plan
totry to sell it or market it. If I do neither of those things, then it should be made public.
 

If, on the other hand, I develop something useful, but not world-breaking, that may still need work, then I will
releaseit into the (internet) world, and hope some other person:
 

1. Finds it useful and saves them time, so they can do other GPL work.
2. Likes it and enhances it (thereby saving me time).

If enough people find such code useful, it may eventually become a 'PgSQL-scale' project, and I'll be very happy.

>Commercialization of code is not a bad thing.  

But you need a pretty good reason to do it!


>Fortunately for Linux, there are enough non-paid programmers working on
>it that GPL is not a problem.

Not to mention the paid programmers...


>Maybe all software will some day have
>enough non-paid programmers so commercial software organizations with
>teams of paid programmers are no longer required.  Maybe not.

I get paid to write software for people. The nature of my work means that unless otherwise specified, I own the
copyright.I get a lot of work because my clients are happy, and I work fast. I work fast because I reuse my own code (I
willnever GPL my own software libraries!). If more commercial organizations pooled their code (eg. via GPL) then I
wouldbe out of a job, and they would save a great deal of money. This won't happen because most organization have
mistakenbeliefs about their 'competitive edge', even when it relates to non-core business.
 

Teams of programmers will always be employed either to support and enhance software (legacy code, or GPL'd stuff), or
todevelop 'proprietry' code (neural-database-thingy), for projects that are too specialised to warrant general interest
(electronicfuel injection systems), or for mission-critical code (heart monitoring machines). Only the first in this
listis GPL'd, but it will be the largest category.
 

We are the factory workers of the new millennium - as such, over time we will probably face more 'piece-work', than new
developmentwork, but we will have work.
 

>Here are some short examples.  I have a Viewsonic 15" digital flat panel
>monitor with ATI XpertLCD card.  Xig has a commercial X server that
>drives it.  XFree86 doesn't support it. 

Usually there is a reason for this. XFree used not to support Diamond Stealth cards, until Diamond made some data
public(they presumably believed the data gave them a 'competitive edge'). By then I had bought another card.
 

>The cost of the X server is
>worth it, because without it, I would be forced to us another display
>device.

If the X-server was $1000, you would have bought another monitor and say 'the price of the monitor was worth it,
becausewithout it I could not run Linux'. Everything is relative; the best solution was that you had to pay no money,
andXFree supported the monitor. Xig may even have made more sales into the Linux market if they made their X Server
GPL.I presume they are a hardware manufacturer? Most people give away their drivers for a very good reason...
 

>The cost of BSDI is well worth it for me, because of the high
>reliability and performance of the OS is well worth the cost.  

I could say the same about Linux. My linux box is substantially more reliable than my NT box. You could at this point
say"Yes, but NT is Microsoft", but my point is that Microsoft is probably the embodiment of the anti-GPL philosophy.
 

>Free
>software is nice, but for me, the cost of commercial software is a
>bargain considering the benefits it provides.  (This doesn't mean I
>don't support open software.  I am a PostgreSQL developer.)

For me the support, flexibility, and reliability of the GPL software I use is substantially superior to the commercial
offerings.

>
>Who do you want to write your heart monitor software?
>

As somebody else said, "someone my family can sue". But in a more serious sense, heart monitoring software is probably
asmall niche that *does* require 'due care' be taken. It will be commercial for a long time to come. Personally I would
preferit be released under a GPL, then 1000's of programmers with heart problems will find the bugs before they kill
someone...


At 22:14 20/06/99 -0400, Clark Evans wrote:

>Thus, although you you have found some noteable exceptions to 
>Bruce's comments, the general thrust of his argument still 
>holds -- if the software distribution market was competitive, 
>companies like RedHat, etc., could not afford to fund open 
>source development. 

I do not agree with this. Red Hat is *not* competing on the basis that it's source is better. The fact that Red Hat is
a'standard' Linux distribution is crucial to it's sales. What the likes of Red Hat use to define themselves is how they
packagethe software (no disrepsect meant): the quality of RPM, the SUPPORT they provide, and the fact that everything
isopen. If 'Blue Hat' came along and wanted to compete, they would not try to out-code Red Hat, they would try to
providebetter installation, support and distribution mechanisms. 
 

>However, RedHat's business is NOT distribution, it is
>standardization.  And this allows them to spend money
>on open source _if_ it is in their best interest.  I would
>argue that it is in their best interest now, but it
>won't be in a few years after they have a fimly 
>established monopoly. 

In fact the *only* way Red Hat can become a monopoly is by 'owning' the code. They may dominate, by force of numbers of
developers,but so long as the GPL applies, they only own the good will they generate and the distribution and support
businessthey establish.
 

This does not seem too unreasonable, but maybe I'm naieve.

>Thus, your exceptions point to a deeper problem with
>open source, rather than positive support for it.

The only real problem for open source is ensuring that the 'reference' copies of the software are not all controlled by
onecompany. Which is another reason why it is in developers interests to continue contributing to open source projects.
Ifno-one contributes, Red Hat will have an effective monopoly.
 

----------------------------------------------------------------
Philip Warner                    |     __---_____
Albatross Consulting Pty. Ltd.   |----/       -  \
(A.C.N. 008 659 498)             |          /(@)   ______---_
Tel: +61-03-5367 7422            |                 _________  \
Fax: +61-03-5367 7430            |                 ___________ |
Http://www.rhyme.com.au          |                /           \|                                |    --________--
PGP key available upon request,  |  /
and from pgp5.ai.mit.edu:11371   |/