Marcus Denker wrote:
> Yes, we need to use package maintainers more. But: Keep in mind that  we 
> come from a status where just everything was ignored and *nothing*  
> happend.

Well, that's only true if you look at the problem with tunnel-vision. If 
you look outside the particular SqueakX.Y-basic.image you will find that 
*lots* of things happened.

But I never disagreed that one possible solution to the problem is that 
you and Stef just take over all the code and do everything. Except that 
it doesn't scale and burns you out. Which was the starting point of this 
discussion.

> e.g. you said that the people who wanted to package the image failed.I
> don't think so: They actually did a lot of chagesets, all of which are
> now im 3.8, but it took over a year to harvest them!  And thus people  
> stopped to put work into that untill the changes they already did get  
> included.

Oh, yes, code got produced. That's not the failure I mean. The failure I 
mean is the failure to find people who feel actively responsible for 
various parts of the image and enable them to do their work to the best 
of their abilities.

> But there are changes that can't be done differently... e.g.  
> monticello: it overrides a method from HTTPSocket, thus "unfixing"  
> everthing we do there. Now
> we moved the poxy into thte preferences (so normal people can set the  
> proxy...), but this means this method changed in a way that the one  
> provided
> by Monticello does not work anymore, but a MC without that override  
> would not work in the old images.

I think these changes *would* have been done differently, if there were 
a maintainer to talk to and to either fix the bug for real (if it was a 
bug) or to work out how to integrate a hook for Monticello (if it were 
that kind of override). To me, the mere existence of that override says 
that we have either a maintenance or a communication problem.

But I would make exceptions for overrides in general since they are evil 
to begin with. As a matter of fact, one of the rules that *I* would have 
for packages that should go into releases is to require them to have no 
overrides (and only very well motivated extensions). That way you know 
you don't need to hack it. And it makes perfect sense for both sides, 
since the package stays clean and the integration becomes easy.

Cheers,
   - Andreas