gts@VIOLET.BERKELEY.EDU (Greg Small) (10/19/87)
= BECKER@HUMBER.BITNET (Bruce Becker) = = I'm naive here, so this question may be off base: = Why not use the NetBios interface and library calls developed around it? = I haven't got a clear idea what the limitatios of Netbios are in terms of = functionality vis-a-vis TCP/IP, so I'm really looking for comment here... = Bruce Becker Humber, College Etobicoke, Ont. Actually Ungermann-Bass has already done this, we have been using it since June 1986. They extended the Netbios interface to include direct access to the TCP, UDP, IP and raw Ethernet. Because these are available at the same time, we run PC Network, telnet, ftp, tn3270 and user applications based on Netbios or TCP/UDP/IP simultaneously. Their 4.2 BSD socket library allows BSD applications to be ported. I asked Greg Minshal, who ported his tn3270 to this interface, why he hasn't mentioned UB's Netbios interface in this group. He said: = minshall@opal.berkeley.edu Fri Oct 9 09:19:16 1987 = = The UB interface is not a bad start. It isn't quite extensive enough, = however, to support all the socket semantics (part of that has to do with = the interface the NIU provides). For example, one needs to be able to say = "YOU assign me a local port number". The interface doesn't support this. = = Also, I think that the INT-side of the interface (ie: the INT server) needs = to keep more state information about connections than the UB interface does. = = Finally (for this), the UB interface is dificient in certain aspects. On = transmits, one needs to be able to say "transfer as much as you can to your = internal [eg, on-board] buffers, and tell me how much was transferred". One = needs to have a service of getting an interrupt whenever some state (ie: = transmit space available, or receive data available) changes. In addition = to an "allocate", one needs to be able to do a "receive as much data as is = currently available, and tell me how much that was" operation. = = I'm also not particularly sure that using the NETBIOS INT is a very good idea. = What if IBM/Microsoft/whoever decides to expand the NETBIOS functions to = overlap the TCP range? = = (On the other hand, I don't see anything wrong with pointing at the UB = interface as a first example; in fact this is probably a good idea.) Greg Small (415)642-5979 Personal Computer Networking & Communications gts@jade.Berkeley.EDU 216 Evans Hall CFC ucbvax!jade!gts University of California, Berkeley, Ca 94720 SPGGTS@UCBCMSA.BITNET
geoff@eagle_snax.UUCP ( R.H. coast near the top) (10/26/87)
In article <8710191858.AA17934@violet.berkeley.edu>, gts@VIOLET.BERKELEY.EDU (Greg Small) writes: > = Why not use the NetBios interface and library calls developed around it? > > Actually Ungermann-Bass has already done this, we have been using it since > June 1986. They extended the Netbios interface to include direct access to > the TCP, UDP, IP and raw Ethernet. Because these are available at the same > time, we run PC Network, telnet, ftp, tn3270 and user applications based on > Netbios or TCP/UDP/IP simultaneously. Their 4.2 BSD socket library allows > BSD applications to be ported. There's been a real lost opportunity here. Back at the beginning of the TCP/IP-on-Netbios standadrization effort I suggested to U-B and Excelan that it would be really good to have a parallel convergence on a standard for talking to a resident TCP/IP system, using a variant of the U-B extended Netbios interface, so that we could get a BINARY interface standard rather than relying on more-or-less compatible socket implementations to give us a rather tenuous source compatibility. Sadly, nobody was willing to commit to it. If we had, we'd all be happily interoperating these days (maybe).... -- Geoff Arnold, Sun Microsystems | "Picture a bright blue ball, East Coast Division (home of PC-NFS) | Spinning, spinning free; UUCP: {ihnp4,decwrl,...}!sun!garnold | Dizzy with possibility... ARPA: garnold@sun.com | Ashes, ashes, all fall down..."