17 Mar
2018
17 Mar
'18
8:22 p.m.
On 17 Mar 2018, at 13:43, Clem Cole <clemc(a)ccc.com> wrote:
Simply out (and for those) that don't want to reads the more details arguments -
please don't try to compare Fortran to C, Pascal, Java, Rust etc. or many other
languages - please do not knock it because you don't need to use it or look down on
those that do use because it helps them. But, instead remember that is in your toolbox,
has been and is an appropriate solution for many problems, and is likely to continue to be
for many years.
[...]
Today's Fortran is not, the language Backus and team at IBM created in the late
1950s. Like English (or 'American English' maybe), it has morphed a bit and
taken ideas from other languages.
Also without wanting to start a war about this, I want to agree strongly with it. I work
somewhere where our main computational tool is a large Fortran program which does things
critical to the security (both economic and defence) of the country I live in. It's
officially in Fortran 90 but I think older chunks of it are probably still in FORTRAN 77
and yet other chunks written to more recent standards.
It's a horrible thing, but it's a horrible thing because it has been written by
scientists rather than people who have software backgrounds, and written & maintained
over something like 30 years or more. In particular it's not horrible because
it's in Fortran: Fortran 90 is a reasonably pleasant language as far as I can see (I
learnt FORTRAN 77 and since I don't work directly on this program I'm not really
familiar enough with the later standards to make a strong statement), and later standards
seem even more pleasant.
We're in the early stages of replacing this program by something which will scale
bette (to, eventually, millions rather than thousands of cores). That program is going to
be written in Fortran (with fairly extensive preprocessing to isolate science code from
details of the implementation), and that's *the right decision*.
--tim