I don't buy that as a valid argument- the same thing could have been said in response to the notion of having classes contain subroutines. "If I am in the debugger looking at subroutine blah(); how can I know which blah() I'm looking at?". I don't think the solution to that problem turned out to be all that bad, and I don't think it matters whether you are dealing with closed or open source.
- les
----- Original Message ----- From: danielv@netvision.net.il To: squeak-dev@lists.squeakfoundation.org Sent: Thursday, October 31, 2002 11:49 AM Subject: Re: An uncomfortable question
I think if we ever have selector name spaces and such (that produce the capability I call "co-existance") there should be no link between the name spaces, versions, and dependencies.
To try and predict and specify in advance which versions of other modules this module will work with seems to me very similar to declaring variable types. I think it's the same sort of decision (I think I know what kind object will come here..), with some of the same costs (maintainance, cost of change rises, opportunities lost).
Anyway, Squeak is very far from being used as a typical OS, and maybe never will. The real question is, does this technology, that will require low level changes, and will require people to let go of some important working assumptions(1), solve a problem we actually have, right now? is that problem that big?
(1) example - "if the debugger stack says I'm in Bla>>blur, and I'm in the Browser, looking at that method, and I haven't changed anything, then I'm definitely looking at what was running when the error happened, because that's all there is". Let go of enough such assumptions and you bring the barrier of entry into coding in such a system high enough, that your community will change in ways you might not like.
Again, the tradeoffs for this decision are very different if you live in a closed source world...
Daniel Vainsencher