[E-Lang] Deadlock-free [was what is good about E?]
Tue, 17 Jul 2001 16:47:19 -0700
> I am very excited about coding in E because I will never be forced to deal
> someone else's multithreading code, to argue with a teammate about the
> multithreading, or to wrap non-blocking APIs on top of blocking APIs to
> that I need to use.
> But I already have five salient features on my list: distributed,
> secure, object oriented, and dynamic, and that is enough for me.
Deadlock-freeness does not belong on the list at this level of detail: it is
part of the subtext of why E is "distributed". My only problem with your
list, Zooko, is that, in a marketing slogan faceoff, JavaSoft can jump up
and down and say, "me too!" to this list (well, except for "dynamic", I
guess, which is the most controversial of your items from a marketing point
of view--lots of people just love that ol'-time static type checking
religion--including me on Tuesdays when I am writing servlets that I can't
load with a debugger). This is why I keep wanting to squeeze a little more
detail into the discussion, even in the marketing faceoff: not merely
distributed, but deadlock free as an example of why it is distributed, not
merely secure but implementing POLA, that self-evident best-practice that,
hmm, you just can't get with firewalls and Unix ACLs (as my elevator pitch
has been evolving lately, I have been shifting towards talking about POLA
not capabilities, POLA is so easy to understand, capabilities happen to be
an implementation strategy; call it "POLA-based security" :-).
POLA not Passwords :-)