Re: Random comments and long memories

[Paul Haahr <haahr@mv.us.adobe.com>]

> I'd like to wish for first-class environments (for EVAL)
> and continuations (call-with-current-continuation).

first class environments (or user accessible hash tables) nearly appeared
in the language, but it started feeling too little like a shell.  probably
just nearsightedness on our parts.  as to eval, the read-eval-print (or,
in the case of a shell, read-eval) loop will become user-replacable soon.

call/cc will probably appear when es becomes properly tail recursive.
i don't expect to see that before the 1.0 release we're planning to
make at the same time as we give the es paper at usenix.

also, as to what you suggested in your previous note about user supplied
parser editing and/or reader functions, we couldn't figure out how to do
that properly and support all the weird shell syntaxes we wanted to support.
(pipes, redirections, etc.)  suggestions are welcome, but we had not planned
on that until the next shell.  (``daughter of es''?)

> Apart from the bug, couldn't we just have extra elements of
> prompt silently ignored, like in rc?  I find it handy to carry
> a third one around.

ok.  sounds reasonable.  i hadn't realized anyone would take advantage
of that.

> but in the course of doing that the
> handler->up is NULL, so bottomhandler gets assigned NULL and when
> vardef tries to re-throw the exception the assertion fails.

i don't think handler->up should ever be NULL---it gets initialized
from bottomhandler, so that just shouldn't be.  i can't reproduce
it here, but that just indicates compiler differences.

--p