G/M, A look through 'current' Perl source reveals several instances of use of the FD_SET/FD_ZERO/FD_ISSET macros, whereas the pp_sselect interface uses 3 bit-mapped 'strings' to identify the sockets to be used by the select() function. Assuming there are no technical constraints, I wish to propose that the IO:select interface to pp_sselect be changed to use 3 'int'(I32) arrays to hold the socket ID's and that pp_sselect use the FD_xxxx macros (as used elsewhere in Perl) to adapt these to select(). This would have the following benefits: 1. IO:select.pm no longer needs to work with bit maps, 2. The interface is consistent with other places in the source, 3. OS's that support FD_XXXX macros but not bit-mapped strings (f.e NetWare) should now be able to work with select(). This is based on the (reasonable) assumption that for all OS, the FD_XXXX macros use the same parameter type returned as a socket ID. Or have I missed something? Regards, NormThread Next