>but I don't know how many people in developing countries will quickly switch to VFP 9;
I am also ambivalent on MS not supporting the newer versions of vfp on older OS.
On the one hand NT4 was a rock solid platform (whereas W2K for a long time was not welcome / considered safe even in banking circles) I standardized all my machines on NT4 for a long time, even if the newer hardware was not totally supported (sound is not really a necessity for work <g>, but USB and FireWire became more important in the last years). And I am furious if I have to tweak some settings to get unneccessary eye candy from sapping processor cycles.
On the other hand: if you have machines with oodles of memory and data crunching jobs, W2K is a great boon just by better using "left over" memory to cache files. Even with the help from Systernal's NT-tools I was never able to squeeze the performance from NT4 up to W2K level, so there are also technical benefits by "not supporting" old OS's.

>nor, how far all this can be trusted to really work.
Rule of thumb: factor between 3 and 5. You also know most of the time enough of the basic parameters like reccount(), fsize()'s of the key expressions and record sizes, so your on-site guesses will be quite accurate. Since you have a property for enabling you get very fine grained control and in theory are always better of. And it works wonders if you "inherit" an application already in this kind of trouble<g>. On existing solutions I'ld now always check this before trying other ways to optimize.

>But, if a filter does work well with a grid - either because reccount() is not too big, or because the previous problems were solved - then I think the filter would be easier to implement.
Same here: while the view/cursor/CA based approach is "cleaner" and safer from accidental corruption, it takes more time and cost. So if your customer is striving for bare bones solutions, these can "scale" now to higher record numbers.

I personally am quite happy that most of my customers have left vfp6SP5 behind, although this was a milestone in stability and possibilites to work (COM based) with other applications. A few of them run vfp8 solutions even on NT4 - having their own IT department, knowing about the risks and having taken responsibility for any of those problems<g>.

Yes it is a "carrot and stick" situation, but staying in vfp6 you also automatically loose YOUR chances for the carrots as well. Having newer versions in your arsenal gives you and your customers more options - you can always drop back (even if you will hate it <g>).

regards

thomas