Open-Source and programming style

Open-Source and programming style

[Tom Moran]

One advantage cited for Open Source Software is that it can be
debugged in parallel by many people.  That would seem to fit the style
of 'code anything, then debug until it works' better than the 'design
it so it works in the first place' style, which seems less amenable to
parallelism.  Comments? 

Re: Open-Source and programming style

[Jerry van Dijk]

Tom Moran <tmoran@bix.com> schreef in artikel
<364d0243.39960214@SantaClara01.news.InterNex.Net>...

> One advantage cited for Open Source Software is that it can be
> debugged in parallel by many people.  That would seem to fit the style
> of 'code anything, then debug until it works' better than the 'design
> it so it works in the first place' style, which seems less amenable to
> parallelism.  Comments? 

Ever seen a design, much less a requirements document for
Open Source Software ?

-- 
-- Jerry van Dijk | Leiden, Holland
-- Team Ada       | email: jdijk@acm.org
-- Ada & Win32: http://stad.dsl.nl/~jvandyk

Re: Open-Source and programming style

[dewar]

In article <01be0ff2$6dd17b60$96a55c8b@aptiva>,
  "Jerry van Dijk" <jvandyk@ibm.net> wrote:
>
> Tom Moran <tmoran@bix.com> schreef in artikel
> <364d0243.39960214@SantaClara01.news.InterNex.Net>...
>
> > One advantage cited for Open Source Software is that it
can be
> > debugged in parallel by many people.  That would seem
to fit the style
> > of 'code anything, then debug until it works' better
than the 'design
> > it so it works in the first place' style, which seems
less amenable to
> > parallelism.  Comments?
>
> Ever seen a design, much less a requirements document for
> Open Source Software ?


I don't think you have any basis for assuming that Open
Source Software is any less well designed or less subject
to formal requirements specification than proprietary
software. It is really a completely orthogonal issue. It
is a marketing decision, not a technical decision to make
your sources open. There are well designed OSS applications
and poor ones, and is the case for proprietary designs.

The notion of lots of people debugging in parallel, and
changing the sources with little discipline is merely one
possible model of OSS development, not a very good one in
my opinion. Frankly I have seen lots of proprietary
development which suffered from the same weakness!

Certainly this model is NOT the model we use for GNAT,
where we very carefully consider design issues, and where
of course there most certainly is a requirements document
(it is called the ADa 95 RM :-)

There are those that are highly critical of the OSS
approach. I usually find they are people who have a big
investment in proprietary software, and who, like Microsoft
in the Halloween document, feel, quite understandably and
quite justifiably, under pressure from the OSS phenomenon!

Robert Dewar
Ada Core Technologies

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own    

Re: Open-Source and programming style

[Jerry van Dijk]

dewar@gnat.com schreef in artikel <72knmb$q79$1@nnrp1.dejanews.com>...

> > Ever seen a design, much less a requirements document for
> > Open Source Software ?
> 
> 
> I don't think you have any basis for assuming that Open
> Source Software is any less well designed or less subject
> to formal requirements specification than proprietary
> software. It is really a completely orthogonal issue.

I would agree in the general case, but my note was a reaction on
Tom's, who had an -IMHO- implicit link to the bazar style of
development. And as I understand it, in this model software
is not so much designed as, well..., grown, so to speak.

> Certainly this model is NOT the model we use for GNAT,
> where we very carefully consider design issues, and where
> of course there most certainly is a requirements document
> (it is called the ADa 95 RM :-)

:-)


-- 
-- Jerry van Dijk | Leiden, Holland
-- Team Ada       | email: jdijk@acm.org
-- Ada & Win32: http://stad.dsl.nl/~jvandyk

Re: Open-Source and programming style

[Tom Moran]

>implicit link to the bazar style of
>development. And as I understand it, in this model software
>is not so much designed as, well..., grown, so to speak.
My original question came after contemplating the "Halloween" memo.
  If debugging time costs as much, in both calendar and man-hours, as
design time, and an extra hour of design can prevent multiple hours of
debugging, then clearly more design/less debugging is the reasonable
way to go.  But if you have an environment where design is expensive
and debugging cheap, then less design/more debug seems clearly the
rational tradeoff.  (That is, after all, what Darwin tells us Mother
Nature has been doing, with rather spectacular success.)
  It's also my understanding that the big OSS projects have been in
situation where the spec was already pretty well understood by many
people (eg, Unix, the Ada LRM) rather than needing specs for a
completely new, never been done before, undertaking.  The latter
seems more in need of a very small design team than the former. 
 

Re: Open-Source and programming style

[Andi Kleen]

In article <364f3bbe.214201@SantaClara01.news.InterNex.Net>,
tmoran@bix.com (Tom Moran) writes:
>   It's also my understanding that the big OSS projects have been in
> situation where the spec was already pretty well understood by many
> people (eg, Unix, the Ada LRM) rather than needing specs for a
> completely new, never been done before, undertaking.  The latter
> seems more in need of a very small design team than the former. 
 
I think there is a confusion of terms here: Open Source Licensing does not
require bazaar style development (an example is GNAT which
is managed cathedral like). On the other hand bazaar style development
is in principle possible for proprietary software, although that is seldom
done because few organisations have the man power needed for it. Because 
of this reason more OSS projects are done bazaar style - simply because
it is possible (enough volunteers available) and well understood. But it 
is not required. 

Regarding your assertion that big Open Source software always have a fixed
spec because they're cloning something: good counter examples are GNU emacs
and PGP (before it went commercial) 

-Andi

Re: Open-Source and programming style

[Corey Minyard]

Andi Kleen <ak@muc.de> writes:

>
> Regarding your assertion that big Open Source software always have a fixed
> spec because they're cloning something: good counter examples are GNU emacs
> and PGP (before it went commercial) 
> 

Even if part of a piece of software is well-understood, it doesn't
mean that the whole thing is.  For instance, for gcc, the front-end
was well defined but the back-end is, well, quite unique.  And quite
powerful, too, once you understand it.  So the front-end design had a
defined specification but the back end is RMS's own scheme.  At least
that's how I understand it, I don't think it was stolen from anywhere.

-- 
Corey Minyard               Internet:  minyard@acm.org
  Work: minyard@nortel.ca       UUCP:  minyard@wf-rch.cirr.com

Re: Open-Source and programming style

[Richard Kenner]

In article <m27lwwrzb2.fsf@wf-rch.cirr.com> minyard@acm.org writes:
>Even if part of a piece of software is well-understood, it doesn't
>mean that the whole thing is.  For instance, for gcc, the front-end
>was well defined but the back-end is, well, quite unique.  And quite
>powerful, too, once you understand it.  So the front-end design had a
>defined specification but the back end is RMS's own scheme.  At least
>that's how I understand it, I don't think it was stolen from anywhere.

Well, it wasn't "stolen" and there were plenty of new concepts in the GCC
back end, but as RMS says in the GCC manual:

    The idea of using RTL and some of the optimization ideas came from the
    program PO written at the University of Arizona by Jack Davidson and
    Christopher Fraser.  See ``Register Allocation and Exhaustive Peephole
    Optimization'', Software Practice and Experience 14 (9), Sept. 1984,
    857-866.

Re: Open-Source and programming style

[Chris Morgan]

Andi Kleen <ak@muc.de> writes:

> Regarding your assertion that big Open Source software always have a fixed
> spec because they're cloning something: good counter examples are GNU emacs
> and PGP (before it went commercial) 

This is true as far as it goes, however RMS made GNU Emacs based on
his experiences with previous Emacsen dating back to his original
version. It does seem then that GPL is a particularly good fit for
software which is well-understood by the people wishing to develop
it. Cloners in the case of some well-known GPL items, but in many
notable cases the original visionaries arrive at GPL after experience
with other schemes such as proprietary or other "freeware"
licenses. Perhaps GPL is the happy-hunting grounds where programs go
to be happy after having served their time being used to wring money
from users in previous incarnations.
-- 
Chris Morgan <mihalis at ix.netcom.com> 
       Home Web Server -  http://mihalis.dyn.ml.org/index.html

Re: Open-Source and programming style

[dewarr]

In article <87ogq848xf.fsf@mihalis.ix.netcom.com>,
  Chris Morgan <mihalis@ix.netcom.com> wrote:
> Andi Kleen <ak@muc.de> writes:
>
> > Regarding your assertion that big Open Source software
always have a fixed
> > spec because they're cloning something: good counter
examples are GNU emacs
> > and PGP (before it went commercial)
>
> This is true as far as it goes, however RMS made GNU
Emacs based on
> his experiences with previous Emacsen dating back to his
original
> version. It does seem then that GPL is a particularly
good fit for
> software which is well-understood by the people wishing
to develop
> it. Cloners in the case of some well-known GPL items, but
in many
> notable cases the original visionaries arrive at GPL
after experience
> with other schemes such as proprietary or other
"freeware"
> licenses. Perhaps GPL is the happy-hunting grounds where
programs go
> to be happy after having served their time being used to
wring money
> from users in previous incarnations.


This last statement by Chris seems completely bogus to me.
I can only think of counterexamples. Linux itself, which is
the culmination of RMS' vision of a freely distributed OS,
EMACS, GCC, GNAT, GDB etc.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own    

Re: Open-Source and programming style

[Chris Morgan]

dewarr@my-dejanews.com writes:

>Perhaps GPL is the happy-hunting grounds where
> programs go
> > to be happy after having served their time being used to
> wring money
> > from users in previous incarnations.
> 
> 
> This last statement by Chris seems completely bogus to me.
> I can only think of counterexamples. Linux itself, which is
> the culmination of RMS' vision of a freely distributed OS,
> EMACS, GCC, GNAT, GDB etc.

It may be down to language here but here goes. Prof Dewar is always
exact in his language and I am a bit more metaphorical, but I don't
agree that my point is completely bogus. Perhaps it would be more
accurate to say GPL is the happy-hunting ground where the experts go
after having done their thing with the proprietary model.

Linux is a reimplementation of posix and hence other *nixes. Most
previous *nixes have been commercial except for the branch rooted at
UCB. Posix is of course the term RMS coined for the IEEE
standardisation of Unix which is mandated by federal procurement
regulations AIUI (remember, NT is Posix compliant!). Thus this
commercially highly significant "thing" for want of a better word has
finally made it to GPL status and I'm very pleased it has. I get the
feeling lots of standard Unix programs and programmers are very
"happy" with Linux too.

GCC is a unique technology (I mean, whoever heard of a portable
code-generator!), but it has often displaced bundled C compilers which
came with commercial Unix. The basic "compile this C code well,
quickly and in a standards-conforming way" property of C compilers is
now a GPLed commodity (just use the right flags). Note I fully
acknowledge the extra, unique and sometimes astonishing features of
gcc. Similarly with GDB.

GNAT is developed by a formidable tribe of ex-proprietary Ada compiler
developers. Yet again something heretofore largely or wholly
proprietary ("ada compiler") is now GPLed and demonstrating leadership
technology.

In all of the above the transfer of actual source from old proprietary
to new GPL components is minimal, however the expertise applied to
produce the latter has been greatly expanded in each and every case by
experts familiar with the former, along with a few geniuses.

Chris
-- 
Chris Morgan <mihalis at ix.netcom.com> 
       Home Web Server -  http://mihalis.dyn.ml.org/index.html

Re: Open-Source and programming style

[Larry Kilgallen]

In article <8790hbgguf.fsf@mihalis.ix.netcom.com>, Chris Morgan <mihalis@ix.netcom.com> writes:

> GCC is a unique technology (I mean, whoever heard of a portable
> code-generator!)

Tartan ?

Larry Kilgallen

Re: Open-Source and programming style

[Jerry van Dijk]

Larry Kilgallen <kilgallen@eisner.decus.org> schreef in artikel
<1998Nov17.083925.1@eisner>...
> In article <8790hbgguf.fsf@mihalis.ix.netcom.com>, Chris Morgan
<mihalis@ix.netcom.com> writes:
> 
> > GCC is a unique technology (I mean, whoever heard of a portable
> > code-generator!)
> 
> Tartan ?
> 
> Larry Kilgallen
> 

Well, one of my fellow workers...

:-)


-- 
-- Jerry van Dijk | Leiden, Holland
-- Team Ada       | email: jdijk@acm.org
-- Ada & Win32: http://stad.dsl.nl/~jvandyk

Re: Open-Source and programming style

[Richard Kenner]

In article <m31zn4wr4n.fsf@fred.muc.de> Andi Kleen <ak@muc.de> writes:
>Regarding your assertion that big Open Source software always have a fixed
>spec because they're cloning something: good counter examples are GNU emacs
>and PGP (before it went commercial) 

Actually, GNU EMACS was a clone of ITS EMACS, though that was also 
Open Software (in a sense) and written by the same person.

Re: Open-Source and programming style

[Robert I. Eachus]

In article <731eqv$9ff$1@news.nyu.edu> kenner@lab.ultra.nyu.edu (Richard Kenner) writes:

 > Actually, GNU EMACS was a clone of ITS EMACS, though that was also 
 > Open Software (in a sense) and written by the same person.

  Where ITS stands for Incompatible Time Sharing.  ITS ran on DEC
hardware (the PDP-6, PDP-10, DEC-10 and DEC-20 family) while CTSS
(Compatible Time Sharing System), also developed at MIT ran originally
on the IBM 7094.
--

					Robert I. Eachus

with Standard_Disclaimer;
use  Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...

Re: Open-Source and programming style

[Andi Kleen]

In article <01be0ff2$6dd17b60$96a55c8b@aptiva>,
"Jerry van Dijk" <jvandyk@ibm.net> writes:
> Tom Moran <tmoran@bix.com> schreef in artikel
> <364d0243.39960214@SantaClara01.news.InterNex.Net>...

>> One advantage cited for Open Source Software is that it can be
>> debugged in parallel by many people.  That would seem to fit the style
>> of 'code anything, then debug until it works' better than the 'design
>> it so it works in the first place' style, which seems less amenable to
>> parallelism.  Comments? 

> Ever seen a design, much less a requirements document for
> Open Source Software ?

Yes. Lots of them actually.

-Andi

Re: Open-Source and programming style

[Jerry van Dijk]

> > Ever seen a design, much less a requirements document for
> > Open Source Software ?
> 
> Yes. Lots of them actually.

Interesting. Can you point to some of them ?

-- 
-- Jerry van Dijk | Leiden, Holland
-- Team Ada       | email: jdijk@acm.org
-- Ada & Win32: http://stad.dsl.nl/~jvandyk

Re: Open-Source and programming style

[Andi Kleen]

In article <01be1008$d828fc20$65615c8b@aptiva>,
"Jerry van Dijk" <jvandyk@ibm.net> writes:
>> > Ever seen a design, much less a requirements document for
>> > Open Source Software ?
>> 
>> Yes. Lots of them actually.

> Interesting. Can you point to some of them ?

Of the OSS projects I followed for example the Linux-PAM ('Pluggable ]
Authentification modules') project was very well spec'ed/designed/documented 
from the beginning. Another example is gcc/egcs, although the requirement/new
design documents usually only apply to part of the source because they're
already working with a very large code base. A lot of OSS projects implement
based on existing specifications, e.g. like Robert pointed out GNAT based on
the ARM, Orbit/MICO (OSS CORBA ORBs) based on the CORBA specification, etc.

There are of course projects too that work more in 'explorative programming
mode', you just can't generalize.

-Andi

Re: Open-Source and programming style

[Richard Kenner]

In article <m3g1blw9dw.fsf@fred.muc.de> Andi Kleen <ak@muc.de> writes:
>Another example is gcc/egcs, although the requirement/new
>design documents usually only apply to part of the source because they're
>already working with a very large code base.

I think gcc/egcs is actually a bad example, especially in terms of
requirement documents.  It's not like people get together and discuss "well,
what should we put in during Q1 99?".  It's much more of a "Hey great!
Somebody just contibuted most of a new FOOBAR optimization pass!  Let's hack
it into working shape!".

In terms of specifications, of course the base compilers are implementing
precisely specificied languages, but the internal interfaces are indeed not
well specified and this lack has been a serious problem with these
development models. 

A serious problem in a volunteer-based project is that people will volunteer
to do those things they find fun and few people find documentation, code
cleanup, and writing specifications to be fun.  The only way these things get
done is if some central controlling person or group says "You want your code
to be used?  Then you're going to have to do the documentaiton and cleanup we
ask."  In the long term, the success or failure of such projects depends on
the success of that sort of coercion.

Re: Open-Source and programming style

[dennison]

In article <364d0243.39960214@SantaClara01.news.InterNex.Net>,
  tmoran@bix.com (Tom Moran) wrote:
> One advantage cited for Open Source Software is that it can be
> debugged in parallel by many people.  That would seem to fit the style
> of 'code anything, then debug until it works' better than the 'design
> it so it works in the first place' style, which seems less amenable to
> parallelism.  Comments?

I notice a lot of people are attacking this statement on the basis that it
implies no (or ad-hoc) design. However, I don't think that is what is being
asked. If what you are getting at is the "worse-is-better" design approach
vs. the "the Right Thing" approach, I'd have to agree. After all, you can't
get the benefits of parrallel debugging until you actually produce something
nominaly useful in the first place. To quote shamelessly from Richard
Gabriel's The Rise of "Worse is Better":

   The lesson to be learned from this is that it is often undesirable
   to go for the right thing first. It is better to get half of the
   right thing available so that it spreads like a virus. Once people
   are hooked on it, take the time to improve it to 90% of the right
   thing.

--
T.E.D.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own    

Re: Open-Source and programming style

[dewarr]

In article <72pfum$h9c$1@nnrp1.dejanews.com>,
  dennison@telepath.com wrote:
> In article
<364d0243.39960214@SantaClara01.news.InterNex.Net>,
>   tmoran@bix.com (Tom Moran) wrote:
> > One advantage cited for Open Source Software is that it
can be
> > debugged in parallel by many people.  That would seem
to fit the style
> > of 'code anything, then debug until it works' better
than the 'design
> > it so it works in the first place' style, which seems
less amenable to
> > parallelism.  Comments?
>
> I notice a lot of people are attacking this statement on
the basis that it
> implies no (or ad-hoc) design. However, I don't think
that is what is being
> asked. If what you are getting at is the
"worse-is-better" design approach
> vs. the "the Right Thing" approach, I'd have to agree.
After all, you can't
> get the benefits of parrallel debugging until you
actually produce something
> nominaly useful in the first place. To quote shamelessly
from Richard
> Gabriel's The Rise of "Worse is Better":
>
>    The lesson to be learned from this is that it is often
undesirable
>    to go for the right thing first. It is better to get
half of the
>    right thing available so that it spreads like a virus.
Once people
>    are hooked on it, take the time to improve it to 90%
of the right
>    thing.



I think that notion is entirely unsuportable, but in any
case it is not how typical GPL'ed software is in fact
developed. The distinction between the so called (very
inept analogy in my view) bazarre and cathedral models
is largely artificial.

The fact of the matter is that most large GPL'ed projects,
including Linux, GCC, EGCS, GDB, GNAT is that they are
very carefully controlled. For example, in the case of
EGCS, the great majority of modifications are made by
Cygnus, and suggestions for changes outside Cygnus are
vetted quite carefully.

Similarly, the mainstream versions of Linux, such as those
from Redhat, are carefully controlled in terms of what goes
into releases.

Sure there are hobbyists and enthusiasts making suggestions
and hacking their own versions, but the viewpoint that
people have of uncontrolled development is in fact quite
bogus in all these cases.

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own