🚀 go-pugleaf

In article <SBB.90Nov6112327@laplace.sun.com> sbb@laplace.sun.com (Steve Byrne) writes:
   What I'm not clear on is whether the test suites should be part of
   the distributions, thereby increasing their already bloated sizes,
   or if they should be separated out, so that you can get just the
   software and not the test suites if you wish.

A minimal test should be a part of the software distribution, just
enough to ensure that it was installed correctly.  Perhaps it could
emit a string that would be useful to identify bug reports.

A full-blown regression suite should be distributed separately from
the software it tests.  Many users are sensitive to the size of a
distribution, affecting both transmission and storage costs, and few
need to ensure (e.g.) language definition conformance in their own
environment.  Increased complexity also implies higher administrative
costs, since there will be more software to maintain; this absorbs
more of a talented programmer's time, which is a real cost of using
freely-available software.  Reducing such monetary costs (space, time,
and overhead) can be seen as helping further the spread of free
software.

   I guess I don't think that separating them is a real good choice
   given the possibility of version skew...

If a user has the need and the level of sophistication to put a full
regression test suite to good use, then {s}he can probably be expected
to be able to identify and acquire the correct version of both the
software and the test suite.

Hey Steve - how 'bout them Reds, huh? :-)