Relevance of Borland Pascal (was: Standard compatibility)

Fri, 4 Apr 1997 20:03:21 +0200 (MET DST)

According to Kevin A. Foss:
> 
> I have to admit I'm in favor of defaulting to ISO standard behavior, but
> that is just my personal preference.  I'm mainly concerned with the last
> comment... is it true that 'thousands of dos-pascal-programs' will be
> ported to linux'? 

Yes.

There is an incredible number of Borland Pascal programmers out there.
Just watch news://comp.lang.pascal.*: most questions, even those posted
to comp.lang.pascal.ansi-iso, concern Borland Pascal.  Some of them
(~50%???) happily change to Delphi and move to Windows 95 without ever
looking out of the (MS-) Window.  Some others (which I would consider the
best ones;) want to try Linux and are looking out for a Pascal compiler.

> I don't see a majority of the DOS-based TP programs
> being all that useful,

Of course not all of that is useful, but enough to make it worth the
effort.

In addition: think of games.  Not useful, but they can make Linux
more attractive.  Many games are written in Borland Pascal for DOS.
I have heared of game authors who would be happy to have a 32-bit Pascal
for DOS.  If they see that there is a compiler which supports Linux as
well, they will give Linux a try.

> or all that many TP programmers perceiving a need to
> port their software to gpc.   Granted there was a lot of good software
> written with TP and which should be ported, but I think your numbers are
> over stated.

I think these numbers are realistic.  (Just what I guess from conversation
with other programmers.)

> The comment also begs the question. is gpc's role to provide a facility for
> porting 10 year old DOS programs or for the creation of new programs?  I
> would think primarily the latter and that decisions on gpc's defaults
> should serve that end.

GPC is for the creation of new programs.  It should provide facilities
to port existing programs (all those --command-line-switches), but this
should not become our primary goal.

This also guided my decision in favour of a default field width of
0: Most programs written today don't sequentially read in columns
of numbers, process them and put out some other columns of numbers.
Today's programs process strings and interact with the user; most numbers
appear in sentences like Frank's "There are 7 items."  Thus GPC's default
behaviour should support the latter one; those of us who need the first
one (including myself, sometimes) are experienced enough to change GPC's
default by configuration, a command-line switch, or even an explicit
field width specification.  For programs intended to be portable the
latter method is the most secure one anyway.

Greetings,

    Peter

Peter Gerwinski (peter@agnes.dida.physik.uni-essen.de)