Hi Jim,

>You (the VFP Team) chose to limit your supported platforms to Win2K and WinXP in the first place, leaving out Win98, Win98SE, WinME and WinNT.

Wrong. We supported VFP8 being installed for development on Win2K and WinXP but we also ran plenty of runtime tests on Win9x and WinNT platforms. Some aspects of the product, such as GDIPlus, simply weren't going to run on Win95 or earlier service packs of WinNT.

>So I'd say you cut your test requirements below 80% right from the start!
>I said at the outset that it is known to be impossible to test 100% of everything. But if you don't consciously list the whole 100% and then consciously decide which of those are to be excluded then how do you know what percentage has actually been tested???
>VFP8 is extremely STABLE. But it must be noted that some of the bugs reported (not affecting "stability" of VFP) ought not to exist and it seems fair to conclude that there was something missing in the testing (planning and/or execution). This **can** suggest that when one starts out with the premise that 100% testing is impossible then 80% is really just some imaginary number determined by some artificial cut-off. I would have more sympathy for your premise if, for instance, you said that 100% testing for critical/important functionality and 80% for less important and 60% for trivial changes was the objective. You could even limit such detailed test requirements to the latest of MS' home and business OSes, with somewhat less acceptable for other older OSes.

No test process is perfect as is no code. I don't know why you keep obsessing on this issue, Jim, when we've been completely upfront about how and why we test what. For every one thing you have a beef with, 10,000 things work reliably.