This issue brings up some fundamental issues: 1. Althought BeOS might not be a "pure" microkernel, by using servers, it is very microkernel-esque. 2. Theory states that it easier to write and *debug* userland code. 3. Theory states that good and stable userland code can put into the kernel once it has been proven. So, it would seem to be that the server approach would be better. When the userland net_server is done and well tested, one would move it kernel space if one needed to do so at this point. However, if Travis builds a good stack in the NewOS kernel, and OBOS can use it, then it would seem a bit silly to put a lot of effort into a net_server implementation. Finally, if the folks who are writing the stacks want to do both irregardless of the other writer(s), then OpenBeOS can only benefit from having more good code available to use (and it would recreate the R5 and BONE codelevels). -- timothy.covell@xxxxxxxxxxxx