Frank Mittelbach wrote:
> Now after this excure in layers ...
> 
> ... coming back to xparse: to a large extent that implentation provided one
> possible incarnation of a document-level user interface modeled after current
> 2e, in other words it was meant to provide a user interface in which you can
> define user commands that look like current 2e commands and offer the same
> kind of possibilities (like a * or optional arguments present by [...])
> 
> It should therefore probably be better named xparse-2e or something like
> that. After all, it is 
> 
>   a) quite possible to envsion more than one document level interface being
>      available in parallel
> 
>   b) a 2e-like interface, while needed to allow supporting 2e documents is not
>      necessarily the best interface for the future
> 
>   c) depending on requirements and environment on that level interfaces might
>      make use of additional tools to manage the input stream
> 
> if we accept that the current xparse is really xparse-2e (whether we call it
> that or not) then one consequence from this for me is that xparse-2e should
> offer the possibility to define commands with a syntax that fits current 2e.

Not quite sure I follow the paragraph!

> I would expect that once we have a clearer picture of how to do the separation
> between layer -1 and layer 0 all this needs rewriting anyway
> 
> I also hope (and expect) that once we are clear on how to write specifications
> for layer 0 properly, that other interfaces for layer -1 will be written, both
> because more than one might be needed and because we need some trials to
> settle on what we want to promote as that standard layer -1 for latex3

As you probably realise, my overall feeling is that for many (most?) end
users, a LaTeX2e-like syntax will remain the best way to use LaTeX
whatever we deliver as LaTeX3. So xparse having a mainly LaTeX2e-like
focus does not worry me too much. I'd say that I think the underlying
idea is a bit more flexible than just forming LaTeX syntax (as we use
abstract concepts such as "optional argument" in preference to more
concrete ones such as "argument delimited by "[" ... "]").

I don't expect to get everything right as a "hole in one", and have said
before that I'd expect LaTeX3 to vary over time (in a somewhat similar
way to ConTeXt) as decisions are revised. There are simply too many
things to consider to hope to decide everything before implementing it.
However, to do that I think we have to have something which is a usable
LaTeX3 (look at how many other not-finished-but-working projects improve
as they gain user feedback, both in the TeX world and elsewhere). To do
that, we have to get various items to "stable" status so that we can
solve other problems (for example, the output routine). We might well
then go back round and look again things like xparse again, but things
will only really progress if, as we've done with expl3, we decide to go
with what we have for the foreseeable future.
-- 
Joseph Wright