[Shootout-list] Reboot

[skaller]

On Sat, 2005-07-30 at 08:35 -0700, Isaac Gouy wrote:
> --- skaller <skaller at users.sourceforge.net> wrote:
> -snip-
> > That is naive. It is likely the scripts won't work and I'll have
> > to make changes, but won't be able to merge them back so they're
> > not lost.
> 
> You could always contribute your changes back to the Shootout.
> You don't need CVS access to do that.

I need CVS write access to do it conveniently.
How do I contribute 'changes'? To contribute
a patch, I need

(a) the modified code
(b) the original code

and then run diff on it. This is very fragile,
and a pain in the rear end: it needs two copies
of the code for a start, and it assumes the original
hasn't changed in the meantime.

CVS takes care of all that automatically.

There are also ways to ensure changes don't stuff
up other people's code, or don't affect the system
without review and approval, such as using a branch.

So if you want collaboration, you have to use
the tools designed to support it.

> Jon, your idea of open source seems to be that you personally have the
> right to add any source to the project. 

BTW: My name is spelled 'John', 'Jon' is Jon Harrops name :)

Your assumption is wrong, however.
I didn't mention 'open source', I was talking about
contributing to the Shootout. Nor did I ask to be
allowed to add 'any source' to the project, just
to have write access to CVS. That doesn't imply
I would add 'any source' to the project -- any more
than you or any other developer would.

> It's the other way around - you personally have the right to take any
> source from the project.

Good idea not try to teach Grandmother how to suck eggs... ;)

> > So what is your point? I don't require your encouragement
> > or permission to work on my own code. I just think it is sad
> > to duplicate effort by forking projects.
> 
> You have a different idea of what a language shootout should be like.

So what? Everyone has a different idea, that is irrelevant.
If we all had the same ideas, there would be no point
in collaborating :)

It is YOU that seems to be assuming that I would do 
what YOU do -- impose my ideas on others without
discussion. In fact I would not do that .. that is
precisely WHY I have argued on this mailing list;
because I believe in establishing a consensus
before coding.

> Instead of trying to force that idea onto this language shootout, take
> what we freely give and make your idea a reality.

I suggest you stop wasting your breath telling
me what I can or should do: I do not need your
advise on this matter.

I suggest instead you think more logically,
and take people at their word. You asked what
I would do with CVS write access .. perhaps
you should have consider the answer.

One thing I would do is make it work on the
AMD64 I have, and then discuss how to incorporate
results into the Shootout from more than one
test machine: making the code more architecture
independent may or may not require minor
source code changes, adding multiple sources
would certainly require some changes.

Similarly, for people running the Shootout
with a different set of available compilers,
the CVS code should not need modification:
but you told me this can be done by hacking
the Makefile.

There is currently one test, admittedly
to be deprecated, where the test result
is plainly wrong and needs to be fixed:
I even entered a bug in the bug tracker ..
but haven't noticed that you have actually
fixed it: I could have done that in 5 seconds,
and repaired most of the test code for some
of the languages (and found out how to repair
it for others). I even published a repaired
version of the Perl code (which gets it wrong too),
and I'm not a Perl programmer. And my version
is simpler as well as actually being correct.
Did you bother to use that? Nope.

I could go on: there are clearly a lot of things
that need to be done.

You are certainly right we seem to have a different
idea of what the Shootout should be: I believe it
should fairly measure properties that truly
indicate performance and programming features,
whereas you don't appear to care at all: at least
two of us have consistently argued that functional
testing is mandatory to achieve this goal and tried
to explain time and again why this is so.

Similarly, I indicated I believe sequential testing
may introduce systematic errors into the results,
and that the tests could be run randomly to be
fairer: and published code that does that right
here on this list. There are some disadvantages
too .. so of course I wouldn't go and rewrite
the CVS to do that unless there was agreement
(except possibly on a branch, to make it easier
for others to try).

I could go on, but it seems pointless: I have a lot
of development experience and have worked in plenty
of situations where I didn't agree with the goals,
but when you join a team you have to live with the
team's idea, try to change people's thinking,
or leave: you don't just impose your will on 
others.

The bottom line is that the admins have the choice
of what kind of changes should be made and I'd
always respect that: if I didn't, it wouldn't
looks so good for me. The whole idea is that
admins can make decisions, resolve conceptual
conflicts, etc .. but the developers can
do some of the work.

Even if I disagreed with you (or Brent) on some issue
it doesn't mean I'd code contrary to a decision you made
(although I might *refrain* from coding :)

After all .. I have my own project and I'd hope
the developers I give access to it wouldn't stuff
me up .. and that project is a good deal bigger
and more complex than the Shootout.

-- 
John Skaller <skaller at users dot sourceforge dot net>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.alioth.debian.org/pipermail/shootout-list/attachments/20050731/62a8be2a/attachment.pgp