18-May-82 06:43:19-PDT,9580;000000000001 Mail-from: ARPANET host BRL rcvd at 18-May-82 0641-PDT UUCP-From: decvax!duke!unc!brl-bmd!tcp-ip Sender: Mike Muuss From: TCP-IP at BRL To: TCP-IP at BRL Date: 18 May 1982 Subject: TCP-IP Digest, Vol 1 #20 Via: Brl-Bmd; 18 May 82 4:40-EDT TCP/IP Digest Tuesday, 18 May 1982 Volume 1 : Issue 20 Today's Topics: BBN VAX TCP Release for 4.1 BSD UNIX Preliminary Stub Gateway Availible for VAX IP Plans for 6502 TCP/IP Implementation "Starting From Scratch" -vs- Front Ends More on Service Specifications ---------------------------------------------------------------------- LIMITED DISTRIBUTION For Research Use Only --- Not for Public Distribution ---------------------------------------------------------------------- Date: 7 May 1982 11:27:03 EDT (Friday) From: Rob Gurwitz Subject: BBN VAX TCP release To: tcp-ip at BRL BBN has developed an implementation of TCP/IP for DEC's VAX(TM) family of processors, that runs under the Berkeley 4.1BSD version of UNIX(TM). The development effort was funded by DARPA. Some important features of the BBN VAX TCP/IP are that it runs in the UNIX kernel for enhanced performance, it is a complete implementation of the TCP and IP protocols, and provides facilities for direct user access to the IP and underlying network protocols. The IP module supports checksums, option interpretation, fragmentation and reassembly, extended internet address support, gateway communication with ICMP, and support of multi-homing (multiple interfaces and addresses on the same or different networks). The TCP supports checksums, sequencing, the ability to pass options through to the IP level, and advanced windowing and adaptive retransmission algorithms. Support is also provided for the User Datagram Protocol (UDP). In addition to the TCP/IP software for the VAX, BBN has developed implementations of the TELNET Virtual Terminal Protocol, File Transfer Protocol (FTP), and Simple Mail Transfer Protocol (SMTP), for use with TCP. These protocols are operated as user level programs. Also provided are network programming support tools, such as network name/address manipulation libraries, status, tracing, and debugging tools. The TCP/IP and higher level protocol software are now available direct from BBN. The software is distributed on a 1600 bpi tar format tape, containing the sources and binaries for a 4 .1BSD UNIX kernel containing the network modifications and the sources and binaries for the higher level protocols and support software. Documentation is provided in the form of a set of UNIX manual pages for the network access device, user programs, and libraries. In addition, a detailed installation document is provided. Device drivers are supplied for the ACC LH/DH-11 IMP interface and the Proteon Assoc. PRONET Local Network Interface. The tape is available for a $300.00 duplication fee to Berkeley 4.1BSD licensees. To order the tape, contact: Ms. Judy Gordon Bolt Beranek and Newman, Inc. 10 Moulton St. Cambridge, MA 02238 617-497-3827 jgordon@bbn-unix You will then receive a copy of the licensing agreement. Tapes will be mailed upon receipt of a completed agreeement and the distribution fee. This tape is supplied as-is to 4.1BSD licensees, with no warranties or support expressed or implied. BBN would be pleased to arrange separate agreements for providing installation assistance and/or software support services, if desired. UNIX is a trademark of Bell Laboratories. VAX is a trademark of Digital Equipment Corporation. ------------------------------ Date: 10 May 1982 20:57:12-EST From: Chris Kent To: tcp-ip at BRL Subject: Preliminary Stub Gateway Hello, Just a quick note to let you all know that I have implemented an 'ad hoc' stub (i.e., non-routing) gateway inside the BBN (Rob Gurwitz) TCP/IP for the VAX. I say 'ad hoc' because the actual specifications of a \real/ stub gateway are still up in the air; the functionality of my code is derived from lengthy discussions with Rob and Dave Mills. The 'gateway' provides rerouting of IP datagrams to and from a local network, and is transparent to user programs (as any gateway should be). My code is installed in the beta-test version of Rob's code -- whenever the final release becomes available, I'll update it. All this will probably be obsoleted by Berkeley's next VMUNIX release; but we need it now, and 4.2 isn't due till at least Fall. I expect there are other people that would like to connect their local nets to the internet, so I'm making this code available for the interim. If an official spec for the stub gateways appears before Berkeley's code, I'll probably update my code to meet it. Cheers, Chris ------------------------------ Date: 20 Apr 1982 0038-PST From: Mark Crispin Reply-To: Admin.MRC at Su-Score Subject: 6502 implementation of TCP/IP To: TCP-IP at Brl Postal-Address: 725 Mariposa Ave. #103; Mountain View, CA 94041 Phone: (415) 497-1407 (Stanford); (415) 968-1052 (residence) I am considering implementing TCP/IP on the 6502 microprocessor. I have a 6502-based system (an ATARI) and I know of at least one other person who would like to run TCP/IP on a 6502 system (an APPLE). I'd like to make sure nobody else is doing this before I start. My implementation (if I do it) will be assembly coded, and will be modular so that all operating system dependent code would be in a separate module with clearly-defined hooks. I am assuming a TCP/IP link to a 6502 system would be on a modem, although I won't preclude some other hardware implementation. I guess what I'd like to know is how much interest there would be for this. I would do a considerably different implementation if it had commercial potential than if it were just something shared by two or three people... ------------------------------ Date: 23 Apr 1982 1444-PST From: PADLIPSKY at Usc-Isi Subject: Staring from Scratch To: wancho at BRL cc: tcp-ip at BRL It occurs to me that rather than start from scratch on an "inboard" implementation of TCP/IP for a new Host type, you might be better off choosing to use an "outboard" implementation (i.e., in a "network front-end" which employs a compact Host-Front End Protocol to attach to the Host rather than limiting your functionality by attaching in an emultated known device fashion). In a recent visit to another of the Old Network Boys, I discovered that there is at least one firm doing such a product at present. As it's not clear to me whether it's even an announced product yet (and knowing that at least one other firm would like an excuse to build such a gadget), and not being able to vouch for their particular H-FP (not meant as a warning; I simply didn't look at it--it was supposed to be a social visit), I won't thrust the name upon you, but would be glad to pass it on if you're at all interested. (If, by the way, somewhere late in your message to the Digest you asked for such info, please excuse my over-scrupulousness; I only looked at the first paragraph or three before deleting the file and then thought of the NFE suggestion later.) (If, for that matter, the references to NFE's and H-FP's are unfamiliar to you and came off rather cryptic, I'd be glad to go on at length on the topic.) cheers, map ------------------------------ Date: 14 Apr 1982 23:05:25-PST FROM: s r kleiman TO: ucbvax!TCP-IP at Brl SUBJECT: Service Specifications In regard to Walt Hass' response: I agree that the response primitive in the three arrow model is virtually useless. In this context, its only value is as a sort of flow control (i.e. L_CONNECT.response means; you can begin sending data primitives now). In fact, I have been lobbying to try and have all responses removed from the spec. However, I don't think this will happen. I disagree with your points about layering. When the serving layer receives an L_CONNECT.request, sets up a connection, and passes an CONNECT.indication and CONNECT.response, the local user layer knows that the SERVING layer has enough resources for the connection. It does not know whether the remote user layer has enough resources, but I don't believe it is the business of the serving layer to tell it this. This is a job for a peer layer protocol between the USER layers. This is my understanding of the intent of the OSI model. As an example, let us assume, for now, the four arrow model and that the serving layer is the link layer. First the link layer receives a L_CONNECT.request and sends a connection setup request to its remote peer. The remote peer then passes an L_CONNECT.indication to its network layer. The link layer is now stopped until the network layer responds. Note that this response MAY involve even higher layers. The link layer has no way to know how long to wait for a response, unless some timing information about the upper layer(s) is built into it. This is undesirable from a layer independence viewpoint. Also note that there is no machinery in both LLC or LAPB to wait longer than T1*N2. Peer protocols, on the other hand, are designed to have this information. END OF TCP-IP DIGEST ********************