>>Sorry, Dragan. Your English is so good that sometimes I forget that you may not know all of these colloqialisms (sp?). Pegged = Categorized.
>
>Well, this makes two of us in "can't live without FoxTools" club :)

Yep, now if I can figure out the new ones in VFP 6.0.:-)

>Er... colloqUialism? Oh, no, here I go again. I've already won me bad reputation in my own language for things like this. Well, you probably meant my _written_ English... you might change your mind once you hear me :)

Local sayings. You might do the same once you hear me. I don't speak English, I speak American.:-)

>>>Just checked VFP6 hhelp - it's still the same: load one at a time, dump 'em all at once. Anyway, I think we've developed a nice practice of using the .dlls only in those functions where they are loaded. So if they are cleared, they're reloaded when needed. We'd just need a neat strategy for picking the right moment to clear. Seems to be I've gone into strategy again.
>>
>>I think that's not all bad (going into strategy). Again, I think it depends on the usage of the DLLs and/or object.
>
>What I actually wanted is (someone) to measure (instead of me, I'm not patient enough with benchmarking) is the "everyone CLEARs its own DLLS right after use" vs "don't clear them until QUIT". The overhead of reloading vs the overhead of memory consumption and swap file usage. Should be tested on two machines with same disks and motherboards, but different memory (say 16M vs 64M). It should also include testing with functions which are in the Win32API (i.e. the kernel, user32, GDI etc) vs functions which reside in optional parts like TAPI or MAPI and such.

The most important thing is to have a standard methodolgy and stick to it.

>Probably this is where Jim Booth would jump in stating that "the apps are slower by horrible 3% in the first case" and the overall difference is 0.027 vs 0.028 seconds for 10000 calls, and we have already spent ten years worth of this difference writing this.

Aw, you're being too hard on JimB.:-)