[Cuis] Why not just kill changesets?
Juan Vuletich
juan at jvuletich.org
Thu Jan 3 07:53:27 CST 2013
Hi Ken,
(inline)
Ken Dickey wrote:
> My take on the core is that it should be like the VM, a set of coherent services which everyone can count on. A baseline and interface.
>
> My experience with Tektronix's Smalltalk is to observe that many people made changes to the core with the result that any attempt to merge results together broke things.
>
> So to extend the stump metaphor, people made limbs that required a stronger heart, lungs, and so forth but when others wanted to use a new limb other limbs did not match up with changed arteries.
>
Yes. The same happens today. The best we can do is to integrate as early
as possible. That's why the trunk was so good for Squeak. In Cuis, we
also want to integrate contributions early, and discuss enough about
changes that could affect compatibility with other ongoing efforts.
It might happen that some people want to do changes that are
irreconcilable with what others are doing. That's the main reason for
forks. Like Tweak, Cuis and Pharo.
> Wow, that was an incredibly bad attempt at an explanation. 8^(
>
> What I should say is that having a single coherent baseline is incredibly important for rapid progress. I don't mind that the baseline itself changes and evolves, but the evolution must had a common, single thread.
>
Agreed.
> I also like building the core from source, rather than just evolving images, but I want to assume and start from the core and not have to assemble it each time.
>
Me too.
> As you can see, I have not yet had my coffee today.
>
> Apologies for the poor communication attempt.
>
> -KenD
>
>
:) I think you made your point.
Cheers,
Juan Vuletich
More information about the Cuis
mailing list