I think the harshest criticism I can muster of interlisp medley is that
so far as my toe dipping is aware, it has lots of similarities to modern
emacs.
(ed 'bubble 'function)
In general cedet-derived emacs modes for whatever language will stub a
function for you. The outermost sexp underlining is nice, and similar to
what vi or emacs and everyone else do with bracket matching. The
lintless coding style, where there is no concept of me controlling
indentation myself is tighter than anything else that occurs to me.
I haven't yet crossed paths with DWIM Do-What-I-Mean (autocorrect?), or
apropos and describe which I think is styled/integrated differently in
interlisp. There's too much to mention that I know I don't know yet, or
I don't know I don't know yet.
I'm working on my mental model that includes both emacs and interlisp
(original conjectures incoming).
Xerox AI Workstations were expensive lisp machines, back when what we
would think of as a good computer was five figures (down from six, for
the PDP-10). The d machines, interlisp-d is what medley interlisp is
like.
In contrast, overlapping time-wise Lucid Inc (I think) took lisp
machines, and emacs in a microcomputer direction (I might be wrong
here). And xemacs and emacs are pretty similar.
Interlisp has a long history of byte-compilation for virtual machines.
Now my hypothesis is that widespread microcomputers of the early 90s
were not strong enough for a virtual machine running something like a
Xerox AI Workstation.
And so most people I know, including my past self use emacs or a
collection of emacs alternatives for and on microcomputers.
However microcomputers got better and better, and now I theorize that
the Xerox AI Workstation environment has been well matched on
microcomputers (a little late-seeming, but maybe I'm just late to the
game).
What's surprising or unsurprising is that modern emacs is convergent or
analogous to interlisp medley, which is why I drew those lines down the
history of xemacs to lucid and interlisp medley to interlisp-d.
I think medley is smaller, tinier than emacs but much less rotted.
Interlisp is not sbcl (CCL ; ECL ; whatever you use). ECL is a little
obscure, patching in common lisp as a C lib but interlisp for me was an
unknown. Well, I think interlisp has a common lisp implementation inside
of it (though its project says its conformance to the standard is a
pretty open question).
I lost a train of thought; due to interlisp's byte-compilation strategy,
once microcomputers were all good enough, to a reasonable extent
interlisp virtual machines were just droppable into unix-like operating
systems.
Interlisp is itself a universe; graphics, ultra-powerful editor
environments, mice and keyboards, once I figure Xnet? out, networking,
fancy revision control, filesystem interlisp programs, strategies and
conventions.
And that universe's culture and knowledge, a portal between AI autumn
and the new spring.
Irresistable. *But* that's all we've currently got. The 80s and right
now, whereas what is normal now is the intervaling sbcl-dominated world
(well, presumably Franz Inc and Lispworks were doing something for their
customers too).
Lots of AI work has been done. Softmax one or two decades ago. Alphago,
deep blue, just right now GPT2 derivatives. Google, Facebook, Windows,
the rise of Nvidia, being a grab bag of timing-based parallel hardware
optimisations.
But interlisp medley doesn't want to be another common lisp compiler
with a C foreign function interface. Interlisp is giant mecha that has
stood still for eons waking up in a new age.