|
|
Tue Mar 2 15:24:57 EST 2010
Backus vs Dijkstra - redux
--
In response my previous post:
Backus vs Dijkstra
I shall say only a few comments, otherwise I might go into an un-
informed rant about functional programming (i.e., prematurely).
I suspect that Dijkstra was not as smitten with functional lan-
guages and programming as Backus was towards the end of his days
[1]. But, [2] was mostly a critique of the speech itself. Howev-
er, it did occur to me that Dijkstra remained unconvinced that
[1] went far enough to extoll the virtues of functional program-
ming versus the supposedly "inelegant" contructs that the von
Neuman architecture imposes onto language design. I remain simi-
larly unconvinced that von Neumann architectures were inherently
bad, inferior to architectures that are supposed to make FP im-
plementations more efficient, or that the "von Neumann bottle-
neck" was terrible thing [1] made it out to be.
[1] was given in the late 70's, and today in 2010 von Neumann
style programming is still alive and well. Functional programming
constructs have certainly made their way into imperative lan-
guages, and mostly where they make sense. While reading [1] (as
well as while reading rants from functional programming
"purists"), I remained unconvinced of two things.
Firstly, functional programming is neat and nice looking, but in
a mathematical way. I get what FP is and why it might seem pretty
and neat. I understand why its benefits are good things. What I
don't get is, what's so bad with imperative programming - or as I
like to see it, practical programming.
Secondly, if imperative languages are starting to incorporate
some of the more practical and useful features borrowed from FP
(hence, making them more like hybrid imperative/functional lan-
guages), why are some people so hell bent on pure functional pro-
gramming? Sure, there are examples of very nice and simple solu-
tions that make imperative examples look as if cavemen wrote
them; but there are a great many practical things that break the
purely functional model of programming.
I am not against FP - it provides a nice view of programming; but
I can't see how it taken to its logical conclusions offers a bet-
ter solution than a hybrid language. The future of languages is
going to be based on augmenting imperative languages with those
features borrowed from FP that produce a significant practical
benefit.
In conclusion, I was a bit turned off by [1]; and clearly it's
been taken up by people who dream of a purely FP utopia. They're
in the great minority, and to be frank, they annoy the crap out
of me. I'm attempting this exercise myself, but I'd be interested
in seein an example of how one might take a recipe for baking a
cake (i.e., an imperative form) and put it into a functional
form. Maybe the result will be very clear and concise, but I
doubt it - perhaps one can publish a book of recipes in function-
al form and see how many house wives and grandmothers flock to
it. A Better Holmes classic it will not be. I am going to think
more about this issue of the von Neumann bottleneck. Sure, [1]
paints a pretty ugly picture of it, but I don't think its an ob-
stacle more than just the nature of how a human being might prac-
tically implement random access to memory.
References
[1] http://www.stanford.edu/class/cs242/readings/backus.pdf
[2] http://www.cs.utexas.edu/users/EWD/transcriptions/EWD06xx/EWD692.html
--
Powered by vee
Copyright © 2006-2010
|
|