Joseph,

 > Thanks to some fast reading by Will Robertson, I have a preliminary TUGBoat
 > draft ready.  I've posted it at
 > 
 > http://www.morningstar2.co.uk/papers/LaTeX3.pdf
 > 
 > Feedback, suggestions, etc. are all welcome.  It is a slightly more detailed
 > version of the first half of my talk.

I like the article (and as I said once it is ready I would like to like or
have a copy for the LaTeX web site).

a couple of comments (in random oder)

 - I already made a remark concerning tlp not being similar to strings (tokens
   aren't chars so token lists aren't strings) I think this is dangerous to
   imply


 - early on you imply that in section 3.1 + 3.2 you explain the <arg-spec> but
   I don't think that you really do it justice.

     *  \def is rather a weird TeX thing so \def:Npn isn't a good example
        really;  perhaps using something with only standard args would be easier
        to grasps at this point.


    *  more importantly, you don't really mention the big strength of the
       arg-spec concept:

          - once you have a baase function, e.g. \foo:nnn
            then any kind of arg manipulation comes for free, eg

               \foo:cnn \foo:noo \foo:nox  what have you

            and if it is not predefined then there is a single unique way to
            obain the missing variant

            rather than explaining that you go for \exp_args:... which are
            meant to be used only to define the variant, ie not to be used
            withing the code really.

      perhaps reshaping or extending that slightly would be an improvement


all in all it is a nice short summary of what is out there right now and what
it might mean to others, so very much appreciated.

regards
frank