Re: I. M. John W. Backus

On Mar 22, 3:28 am, "Tony D" <tonyisyour..._at_netscape.net> wrote:
> On Mar 21, 11:57 pm, Bob Badour <bbad..._at_pei.sympatico.ca> wrote:
>
> > It's odd that the obit got things so wrong. Fortran is a monstrosity --
> > one of those abominable things that was just good enough.
>
> I seem to remember a quote-oid from Backus along the lines of, "we
> didn't realise we were designing a language - we were building a
> compiler", but I can't remember the exact attribution now.
>
> > It's greatest achievement was to show how not to write a compiler.
>
> I disagree - its greatest achievement was to prove to a lot of people
> that compilers were *possible*. That there was a, to modern eyes,
> unacceptable "bleed" of machine features of the 704 into Fortran's
> design was a consequence of trying to prove the point. Unfortunately,
> the proof of concept was treated as the state of the art by too many
> others for far too long. Hmm, that sounds familiar from somewhere ...
>
> > Backus more than redeemed himself a few years later by learning
> > from fortran how to write a compiler when he and Peter Naur came
> > up with BNF.
>
> Hrmmmr... BNF came about because Backus and Naur both thought they
> understood a section of one of the original reports on the IAL
> perfectly, but it turned out they had different ideas about what it
> meant. Backus presented Backus "Normal" Form, and he and Naur worked
> on it more to come up with Backus Naur Form.
>
> For me, Backus' real crowning achievement was his 1977 Turing Award
> paper on FP, here -http://www.stanford.edu/class/cs242/readings/backus.pdf
>
> > Fortran should be a side-note to BNF and not vice versa.
>
> > (BNF is a very big contribution--more than enough to share.)
>
> I find myself fairly forgiving of Fortran, because of its
> "originating" status. Was it better than Speedcode or similar
> systems ? Maybe, maybe not - but (much like SQL later) because it came
> from IBM it got traction and proved an important point. (I'm less
> forgiving of C because by that time *they should have known better*.)
>
> These days, BNF seems kind of trivial, and compiler usage is taken as
> a given for the vast majority of cases. It's kind of hard to imagine
> things before either of those cases, but that's where Backus was
> working, and his contributions are immense. And to go from essentially
> bare metal on to FP ...

I have found BNF immensely useful - decades after its invention - so I believe much praise is deserved for its development. In fact the ability to convert a valid BNF specification to working code via the C+ + boost::spirit library has saved myself an incredible amount of time. If nothing else it forces the user to design methodically and with some sort of formal rigour.

Fortran has had little relevance to my own work, but I can see the impact it would have had in its time (however monstrous it was). It seems to have been the first step on an iterative process for language design, which we are still very much at the beginning of. As such I believe it deserves recognition whatever its flaws - if just for showing what would be possible. Received on Thu Mar 22 2007 - 10:36:44 CET