🚀 go-pugleaf

In article <1416@sugar.UUCP> peter@sugar.UUCP (Peter da Silva) writes:
<Writing system software in Fortran involves encapsulating ALL the system
<calls and changing them every time you switch to a new O/S, or putting up
<with the slowness of Fortran's I/O library. The Software Tools package
<does the former but provides a slow version that uses the latter approach
<to get the first cut working quickly.

You can get fast IO through a FORTRAN IO library. You just have to do
more work than doing "normal" FORTRAN IO, and have a well-written
compiler/library system.

<C specifies an interface (stream character files) that can be implemented
<fairly painlessly on a wide variety of machines, provides most of the stuff
<that high level system software (compilers, filters, editors, etc...) need,
<and is efficient enough that only extremely weird systems encourage programmers
<to roll their own. Could microEmacs have been written in (say) Fortran and
<still have as much performance and portability?

While the streams abstraction in C is a win, it's not the only
language that has it. And that's not what makes microEmacs portable.
MicroEmacs has an abstract IO level internal to itself - for taking
care of things like OS-dependent file names, error conditions,
end-of-file and end-of-line markers. The same things would work in
FORTRAN. The C stream abstraction merely makes doing a port easier -
there's more likely to be a version you can crib from. On the other
hand, doing all those linked lists in FORTRAN would be a *real* pain.

	<mike
--
When all our dreams lay deformed and dead		Mike Meyer
We'll be two radioactive dancers			mwm@berkeley.edu
Spinning in different directions			ucbvax!mwm
And my love for you will be reduced to powder		mwm@ucbjade.BITNET