Responding to Marcus Overhagen: > Another issue is the decision of the naming of all variables in a > class. > I think it was decided to prefix all with an f, thats not bad, as > long as no one > currently using BeOs also has the same variable name in a derived > class. > I think this risk is acceptable, but probably another prefix like _f > would be > better. That's a good point, although it would "only" affect source compatibility, not binary compatibility - not nice but simply solvable. > The same thing is with endianness issues. If you read an int32 from a > file, No, that's definitely not the same! The first prevents compiling which is unfortunate but easily fixable, the latter prevents the application from working correctly which is much harder to find and fix. I would not make it a requirement to add the _EXPORT label to every exported class, but I would make it a requirement to be endian aware, no matter if it's currently planned to support different platforms or not. Fixing such problems really creates much of unneeded work. Adios... Axel.