Path: utzoo!utgpu!water!watmath!clyde!rutgers!sri-spam!ames!lll-lcc!well!pokey
From: pokey@well.UUCP (Jef Poskanzer)
Newsgroups: comp.lang.misc
Subject: Re: Discussions of languages (Was: Re: Modern langauges)
Message-ID: <5082@well.UUCP>
Date: 26 Jan 88 21:32:10 GMT
References: <1416@sugar.UUCP>
Organization: Paratheo-Anametamystikhood Of Eris Esoteric, Ada Lovelace Cabal
Lines: 30

In the referenced message, peter@sugar.UUCP (Peter da Silva) wrote:
}                   Could microEmacs have been written in (say) Fortran and
}still have as much performance and portability?

Of course.  Don't be silly.  In 1980 I wrote a non-extensible EMACS clone
in Ratfor.  Performance was pretty good, considering it was originally
running on a Modcomp -- ~0.1 sec to update the screen.  Almost as soon as
I got it working, I ported it to VMS and Unix with no problems at all.
A few years later I ported it to TOPS-20 in only two days, and that included
getting the Software Tools bootstrapped.

The reason it was so easily ported is that I used modular design / information
hiding.  It was structured as a collection of library packages, each depending
on lower-level packages.  For instance, at the lowest level of the I/O tree
was a simple raw character I/O package.  On top of that I layered a
cursor-control package (which also used a termcap-file-reading package), and
using the cursor-control package I wrote an optimal screen-update package.
Porting that entire tree to a new OS meant re-writing three routines in the
raw I/O package.  Buffer management was handled similarly, so that only
the lowest level was machine-dependent.

A language that enforces information hiding is nice -- it gives an added
sense of security -- but it's not necessary.  Languages don't make good
programs, good programmers make good programs.
---
Jef

              Jef Poskanzer   jef@lbl-rtsg.arpa   ...well!pokey
When someone says "I want a programming language in which I need only say what
                      I wish done," give him a lollipop.