Frank Mittelbach writes:
 > I think Richard Walker asked whether or not the Team thinks that
 > concepts of modules are relevant for LaTeX3.

Did I?!  No, I wondered about the consequences of using long names.
David answered that one.

Hans Aberg writes:
 >   I just want to point out that the way I worded the modules, it does not
 > exclude Frank's idea of a low level naming convention:
 >
 >   A module would itself define how it uses its names. So a module like \tex
 > could define that all its names should be interpreted directly, that is,
 > \tex/foo is executed directly.

This is what I was assuming too.  So my (our?) counter-proposal has
been that macros be named:

  \hierarchical/path/to/module/perhaps_with_underscores/macro_name:argspec

(or according to Hans - I need more convincing):

  \hierarchical/path/to/module/perhaps_with_underscores/macro_name:argspec:

with / _ : all having catcode 11.  No fancy tricks with
\csname...\endcsname that slow things down by a factor of 50 or more.
:-)

Even if LaTeX itself doesn't take advantage of this convention (by
Hans's <...> notation or otherwise), doc.sty/docstrip, Emacs,
lacheck/chkTeX, etc. etc. might.  I'd like to push forward my earlier
idea of docstrip writing out a file containing a list of macros defined,
macro documentation or whatever, in a form that can be easily used by
smart editors.

To clarify the difference once more: in L3PL as it stands there is
IMHO some confusion caused by using _ to separate module and
description while allowing _ in the middle of the description - and
then not allowing _ in the middle of the module name.  I argue that
the confusion should be removed, preferably by using / to separate
module and description, and allowing _ in the middle of module and
description names.  This also means that there can be a hierarchy of
modules.  As David points out, there are no space and time penalties
(only in the size and handling of latex.ltx plus extra space in the
format for one copy of the longer names).

So instead of e.g. \chk_var_or_const:N we have \chk/var_or_const:N.
I think the latter is also easier to read.

The c, l, and p prefixes could be stuck on the front: thus
\c/iow/comment_char and \l/tlp/testa instead of \c_iow_comment_char
and \l_testa_tlp.  (Hmm . . . this needs some more thought.)

 >   So by the time LaTeX3 is out, and is starting to become accepted, which I
 > will gather will take a few years, those memory concerns will no longer be
 > relevant.

Deep breath . . . are we no better than Microsoft?  A decent TeX
installation is hungry enough as it is.  No more please.

 >   Speed perhaps: If a document can be processed in less than one second
 > instead of ten, that will always be a great advantage.

It is a sobering experience to run any of DEK's stuff (articles etc.)
through plain TeX.  If only LaTeX ran that fast . . . .

Richard.