Jose writes > > > > Is Jose's approach of having an empty interface going to be > > > > accepted? > > > > > > All I've seen so far are a number of "I could live with this" > > > statements, but no strong support for either option. > > > > The main hesitation on my part is that you want to identify > badly setup build > > files as early as possible - preferrably in the parser or > some such. However > > as the interface a class implements can only be tested for > in the very bowels > > of the implementation I am very hesitent to adopt it. I > would prefer a > > manifest entry or flag in type library descriptor or > something like that if > > this was the path we chose to go. > > > > I have no problem in principle with that. As long as we can > eliminate all traces of hardcoded tasks from the parser I am happy. > > This would also mean changes in the current way we install > tasks/types today (i.e., property files), since e need to > pass more than just a name-value pair. We could move to > something like my <antlib> right away.
I would obviously like to see that. Any chance we could get that in a 1.5 release like I originally proposed? The vote I tried to start never ended in any actual votes, just a lot of discussion until Jose sent out his implementation patch. Let's try another vote proposal. Jose - do you want to write up what you did or should I resurrect the old proposal? Cheers, Tim -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
