basile@soleil.cea.fr (Basile STARYNKEVITCH) (06/05/91)
1) RPC sources are publicly available (send a "help" message to archive-server@rice.edu for details). 2) Every NFS implementation uses RPC above UDP 3) Possibly, NFS implementations have a different RPC lib, but they are supposedly compatible 4) as far as I know RPC libraries are therefore compatible (both protocol and source API levels). 5) RPC also exists above TCP - e.g. for long arguments calls (more than 8kbytes in XDR). However, note that RPC is a Remote Procedure Call protocol: thsi means that during a RPC call, the client is waiting for the server's answer. If you need both machine working at the same time, plain RPC is not for you. But you could still use the XDR (eXternal Data Representation) package inside RPC to exchange portable data between different machines (Sun, Vaxen, PC..) . Better, starting from the RPC/XDR sources, you could use the xdrstream struff (which handles XDR messages exchanges above TCP) to implement a (event) message oriented protocol (similar in spirit to X11 protocol..) if this is what you want. To summarize: if you just want a true RPC paradigm, just use RPC. If you want an event driven message facility, you could use XDR.. You could also look at isis (from ken@cs.cornell.edu (Ken Birman) which adresses similar problems (or ASN1/X409 from OSI?) but I don' know them well. AS far as I know, all these (RPCXDR, ISIS, X409) are above a transport level (such as TCP/IP, sockets, TLI..). Basile STARYNKEVITCH Commissariat a l'Energie Atomique DMT/SERMA CE. Saclay bat470 91191 GIF/YVETTE CEDEX France email: basile@soleil.saclay.cea.fr