> > I still think that the model I proposed is best, as we dont need a
> > per-OS assembly to be compiled.
> 
> But you need a per-OS shared library and the code will be slower at
> runtime. So, the two choices are:
> 
> 1) per-OS assembly
>       * faster
>       * smaller
>       * easier to deploy
> 2) OS-indep assembly + OS-dependent shared library
>       * slower
>       * bigger
>       * one more thing to deploy
> 
> The choice looks like a no brainer to me:-)

You also have to consider that you will need to wrap structures, for
example, lets take the `stat' structure.   That is much easier to
implement by using the scheme I suggested.

If we want to have a shared-library-less implementation the parser for
structures will probably be pretty complex to generate a c# version of
it that matches the Unix version of it.

Your turn ;)

Miguel.

_______________________________________________
Mono-list maillist  -  [EMAIL PROTECTED]
http://lists.ximian.com/mailman/listinfo/mono-list

Reply via email to