Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!iuvax!maytag!watstat!dmurdoch
From: dmurdoch@watstat.waterloo.edu (Duncan Murdoch)
Newsgroups: comp.lang.pascal
Subject: Re: old .TPUs under TP5.5???
Keywords: converter?
Message-ID: <1573@maytag.waterloo.edu>
Date: 25 Feb 90 15:17:39 GMT
References: <19557@mephisto.UUCP> <1968@bucket.UUCP>
Sender: daemon@maytag.waterloo.edu
Reply-To: dmurdoch@watstat.waterloo.edu (Duncan Murdoch)
Organization: U. of Waterloo, Ontario
Lines: 27

In article <1968@bucket.UUCP> leonard@bucket.UUCP (Leonard Erickson) writes:
>scott@boa.gatech.edu writes:
>
>From what I've been told, the guy "in charge" of Turbo Pascal refuses to
>commit to any kind of "standard" for unit structure. Apparently he is 
>paranoid about the possibility that if he commits to one he may later need
>to extend it in an incompatible manner. So he feels free to change the
>structure any way he pleases between versions (BTW, TP5.5 units say that
>they are TP6 units...)

That may be true, but even if the structure hadn't changed, or had changed
in an upwardly compatible manner, all 4.0 and 5.0 .TPU files would be useless
with TP 5.5.  

The reason is basic to the design of the TPU:  anytime you change
the interface to a unit, any unit that depends on it must be recompiled.  (This
is because external references are by location in the referenced .TPU file, 
not by named reference.)  Since the interface to the SYSTEM unit has changed 
with each revision, and every unit depends on the SYSTEM unit, every .TPU file 
becomes obsolete with each release of a new version of TP.  

This doesn't seem likely to change.  So why should Borland bother to stabilize 
the format?

And why would anybody bother with a .TPU file if they didn't have the source?

Duncan Murdoch