At 3:41 AM -0400 7/14/99, Peter Smet wrote:

[snip]
>
>Think Bazaar, not Cathedral. I agree that the public release of Squeak
>should not have dangerous or unproven code in it. On the other hand, a
>cutting edge version should have experimental code in it. If you would just
>release Squeak with ANY exception framework then you might be surprised at
>how well this works. Suddenly a whole new set of eyes see and use the code.

I don't get it. Why does Squeak Central have to put the exception framework
into a Squeak release for the Bazaar model to kick in? You can file in (up
to) three exception frameworks right now. If *more* that filing one in is
required to make a difference, then we're back to shifting work onto Squeak
Central. IMHO, if they're going to do the work, they should get to decide
what they want to do :)

>If its' utter garbage and crashes the image, then it wont last a day. People
>will just remove the code.

Well, removing code tends to be a touch more difficult than adding. Again,
I don't see a fundemental difference. If it crashes the image, why would we
want in a "official" release in the first place?

> If it has problems, then everyone can see them
>and attack them in parallel. If other code shows more potential, people will
>put that up instead.

And this is to be distinguished from the current state of affairs...how?

> The more people that look at and use the code, the
>quicker the bugs will show up. The point is, you need to have something
>there for people to focus on and improve, even if it is embryonic. This will
>also help the development of teamwork instead of everyone doing it all on
>their own (for example two versions of OS events, +/- three versions of
>exceptions, x versions of Linux sound support).

But above you wrote that folks will be attacking the bugs (and one presumes
the lacks of functionality) "in parallel", not to mention throughing out
code, adding new code, etc. Kinda like, eh, *right now*! The only
difference is that we haven't got a crashing central release.

I take three exception frameworks (which have some mutual influence) to be
a healthy sign *so long as* they flourish *before* integration into the
central release. Once one becomes "mainlined" it becomes subject to
backward compatibility and there is a corresponding responsibility not to
gratuitously change it. That suggests that a more conservative attitude is
warrented.

(I'll just add that I have qualms about the Cathedral/Bazaar analysis, and,
in particular, I think the "with enough eyes, all bugs are shallow" to be
patantly false. Given *these* assumptions, I'm afraid I'm not as sanguine
that your proposed changes will result in a worthwhile leap in
productivity.)

Cheers,
Bijan Parsia.