On 08.01.2010 02:07, Schwab,Wilhelm K wrote:
Henry,
No argument, your solution is simpler. However, we do not have announcements like this now, and it appears that the poor state of weak collections in Squeak/Pharo is standing in the way of a clean implementation. My vote is to tackle the weaklings, at least until they work well enough to support what we should have in this case. Otherwise, we kick the weak can down the road as has been done all along rather than fixing it, AND we end up with a sub-standard announcments-based framework.
Let's do it right. We will then have an evolving/improving weak collection and something in the image that puts some stress on same. If I'm missing something, please enlighten me.
Bill
Weak collections does not stand in the way of making the use case of unregistering dynamic registrations manually easier. They stand in the way of the use case of unregistrating registeration automatically. Here's what you're missing: No matter how you implement the automatic unregistering, it will be nondetermenistic, for the exact same reasons finalization is nondeterministic. Thus, if you require deterministic unregistration, you have to do it manually, and the use case of making that easier to do have value in its own right. Cheers, Henry