On Fri, 2003-04-25 at 22:27, Matthew Palmer wrote:
> Except that it's typically a lot easier to work out where a program has been
> incompatibly modified ("oops, compile error, damn, the API changed") than a
> standard has been modified. The use of 'diff' notwithstanding.
Well, when you modify a program enough to cause a compile error, sure
it's easy to spot. It's also easy to spot when someone modifies a
standard by renaming all the sections and changing the terminology.
If you really think that its a lot easier to note where a program has
been incompatibly modified, please find all the places where NCSA Mosiac
has been incompatibly (with the relevant standards) been modified to
produce Internet Exploder. Or, if you content that source must be
available, please confirm that GCC-3.2 -> GCC-3.3 won't cause any
regressions in compatibility with the C++ standard, the various
microprocessor standards, etc. The FSF, I'm sure, will be most grateful
for that work.
It's not easy to find program modifications that don't show up as
compile- or link-time errors, even with the use of diff. Programs can be
quite long, and often changes don't change behavior (e.g.,
optimizations). An even larger subset of changes don't break
compatibility.
Attachment:
signature.asc
Description: This is a digitally signed message part