Re: Questions for a language designer

[andrew cooke <address@hidden>]

I wonder if the most interesting languages (those that do something
new) take a list like this and try to find some abstraction that
allows apparently conflicting choices to be viewed as two applications
of a single idea.  So it might be best to treat the list not as
something that directly helps the langauge designer, but as an aid to
identifying where progress is most needed.

The existence of Lisp/Scheme can also be something of a deterrent to
this approach, because it already provides a fairly uniform basis, but
that also gives a good starting point - how can it be improved?  Apart
from type systems (don't want to go there again, please) there are at
least a couple of ideas that seem obvious to me - introducing ideas
into the language from logical programming (Oz/Mozart does this to
good effect) and integrating the environment (Smalltalk, and Lisp in
the golden days of Lisp Machines, I gather).  Other areas, which I
know even less about, include advances in security (capabilities) and
distributed/parallel computing.

This is all a bit naive (apologies for knowing so little - maybe Curl,
for example, already does much of the above), I guess, but I imagine
the document that is being proposed as a graph.  The nodes are
"implementation level decisions" and the arcs represent dependencies
between decisions (ie they connect decisions that apparently
conflict).

The arcs are more important than the nodes.

Would anyone care to identify arcs that will become important in the
future?

Andrew

-- 
http://www.acooke.org