Am 26.10.18 um 08:18 schrieb Will Robertson:
> In general, for your purposes it’s good to “abuse” xtemplate where
> you need to so that we can see what needs to change.
> 
> While we designed xtemplate with fresh eyes a number of years ago
> now, we still think it has a few deficiencies in its model so I
> wouldn’t be surprised if run into areas that it doesn’t work
> perfectly for.

to add to Will's comments, way back I tried myself as using xtemplate 
for that kind of purpose too and it might be helpful to to have a look 
at what I have done back then and see if that fits your thinking or 
gives you some other ideas.

The code from that is in

https://github.com/latex3/latex3/tree/master/xpackages/xfrontm

and I just checked (even the documentation doesn't compile without an 
error -- but that can be scrolled past, though I guess it should be 
corrected). I would assume that otherwise it still works or could be 
made to work but just looking at the .dtx file is good enough to 
understand my thinking from that time.

So I think it is worth a look. What I have done basically is to "abuse" 
the template approach to provide new declaration methods that internally 
use a template, so it is kind of one way templates and  and that already 
is telling that something is not quite right or missing.

My long-term thoughts is that there is a still a big gap in that you 
need a way for local customization of of elements (for which templates 
provide the right kind of abstraction more or less) and the 
configuration of relationships for which you need something totally 
different --- and in my believe what we termed LDB way back and which is 
currently not available except for some old code that need adaption.
My architecture vision talk in India gives a good overview on that

https://www.latex-project.org/publications/indexbyyear/2011/

But given that it is not there, see what can be done reasonably with 
current concept and what requires gymnastics that cry for something else 
is extremely valuable.

frank