On 01/09/2009, at 5:20 AM, Joseph Wright wrote: [Snip something silly I said about writing template-alt-alt.] > I'd point out that most of template-alt was about trying to understand > the code in template (which is not easy to read at all). I'd hope that > I've done enough with template-alt to enable discussion of the ideas > without needing too much more coding until a clear idea emerges of > what > is wanted. Sure, I wouldn't want to spend time re-writing code that didn't need it :) I guess I hadn't thought through my intentions; I think some more comparisons between the different approaches is warranted, especially to convince us all that changes from template's interface, in the first place, is the right thing to do. No point fixing what isn't broken. (Although I think we agree that there are improvements to be made.) > (I'd add that at the moment I prefer the idea of using a > keyval approach as I like the "symmetry" of both defining and using in > the same style. But if it turns out this is not the best approach, > then > I don't see too much difficultly in writing the code to do what Lars > suggests.) Just to clarify, are you happy with the collections/type/template/ instance "hierarchy" in template? Barring some improvements we've discussed such as editing instances and template defaults, I think the broad concept works. When it comes down to it, it matters less what syntax we use for defining templates and more what the template setup allows us to do. For this reason I think it would be good to discuss what template-alt provides over template (such as 'implicit' variable setting and multiple-choice keys) and whether those ideas would work in template itself. More on this later, I hope. Will