_____________________________________________________________________________ news@ftp The FTP Software (R) Newsletter Volume III, Number 2 June 1990 _____________________________________________________________________________ ____________________________________ Since the full multi-tasking capability of OS/2 is used, you can Inside... leave an FTP server up in one window, keep a tape backup operating Item Page in another window, and open multiple New Product News 2, 3 Telnet sessions in other windows. Presentation Manager TM is used to Masthead 2 control these windows, and a mouse click or keyboard command moves you Technical Support 4 from one application to another. Internetworking News 7 With multi-tasking and servers, the PC becomes more like a Corporate Notes 14 minicomputer. Instead of the old ____________________________________ client-server model for PC networking, a peer-to-peer model applies. New Release of PC/TCP(R) for OS/2(R) We also include several new by Bruce Campbell utilities for your mailing pleasure. The pcmail utility simplifies mail On April 4, 1990, four years to reading by allowing your OS/2 PC to the day after the first release of connect to a mail server and then PC/TCP, FTP Soft-ware released process new messages as a batch PC/TCP for OS/2. PC/TCP for OS/2 process. Mail is read, saved, provides the link between OS/2-based replied to, forwarded, and deleted PCs and everything from micros to off-line, saving your mail gateway's supercomputers. It is a complete CPU cycles for more important tasks. implementation of the TCP/IP POP2 and POP3 make sending mail protocol suite and provides the easier by batching outgoing mail, expected client utilities for file thus permitting mail to be composed transfer, terminal emulation, offline. We include unsupported electronic mail, remote command public domain versions of the execution, remote backup, printing, Berkeley UNIX(R) servers for these and network information. clients, so you can get a complete (though unsupported) mail system In addition, servers are included from FTP Software. for the FTP, Telnet, Rexec, Rsh, and Tar protocols, as well as a simple The PC/TCP Development Kit for route server. This means that you OS/2 consists of a complete 4.3 BSD can transfer files directly from an socket library. This socket library OS/2 PC (also true with our PC/TCP is identical to the PC/TCP for DOS for DOS TM product), execute OS/2 socket library, in order to make commands from a remote host, use an porting applications easier. OS/2 PC as a tape backup server, and Libraries for FTP, Telnet, RPC, and use the OS/2 PC as a low performance the SUN extensions will be added in router. the near future. An interface that continued ... _____________________________________________________________________________ June 1990 2 ____________________________________ IBM, PC AT, PS/2 and OS/2 are news@ftp registered trademarks, and Presentation Manager and DOS are Volume III, Number 2 June 1990 trademarks of International Business Machines Corporation. Published by FTP Software, Inc. NetWare is a trademark of Novell, Editor: Nancy Connor Inc. NFS is a trademark of Sun Communications pertaining to this Microsystems, Inc. newsletter should be directed to: Reflection is a registered trademark of Walker, Richer & Quinn. news@ftp SmarTerm and Persoft are registered FTP Software, Inc. trademarks of Persoft, Inc. 26 Princess Street TEEMTALK is a trademark of Pericom. Wakefield, MA 01880-3004 Tektronix is a registered trademark Phone: (617) 246-0900 of Tektronix, Inc. FAX: (617) 246-0901 UNIX is a registered trademark of Electronic mail: news@ftp.com AT&T Bell Laboratories. ____________________________________ FTP Software welcomes suggestions and contributions from readers. will allow applications to be ported Submissions for the next issue of from the DOS environment to the OS/2 the newsletter must be received no environment will also be provided. later than August 30, 1990. We are working on several (C) 1988, 1989, 1990 by FTP extensions to the product. An NFS Software, Inc. Permission to use, TM client and server are in the copy, modify, and distribute this works, and NetBIOS is under publication for any purpose and development. Other applications are without fee is hereby granted, being reviewed for inclusion as provided that this copyright and well. permission notice appear on all copies, the name of FTP Software, Inc. not be used in advertising or Release of PC/LPD for DOS publicity pertaining to distribution of the material without specific Printer ports for large, multi- prior permission, and notice be user machines are expensive and given that copying and distribution computing cycles can be better used is by permission of FTP Software, for other applications. Proprietary Inc. servers such as NetWare TM and LAN Manager work only with DOS machines PC/TCP, LANWatch and FTP Software running the proprietary client. If are registered trademarks of FTP you have a multi-user host which Software, Inc. runs an operating system other than ConneXions -- The Interoperability Berkeley UNIX, such as System V Report is a trademark of Interop, UNIX, there are no standard ways to Inc. do print sharing. To help users DEC, VAX and VT are registered solve these (and other) problems, trademarks of Digital Equipment FTP Software is now selling PC/LPD - Corporation - an inexpensive print server Ethernet is a trademark of Xerox package for the IBM(R) PC and Corporation. compatibles. Grafpoint is a trademark of Grafpoint. With PC/LPD, you no longer have to use an expensive multi-user host to _____________________________________________________________________________ June 1990 3 get complete print services. Take implementation of lpr or a compliant an old PC, connect printers, and the utility. entire network can share this resource. If printer needs vary, a dot matrix printer can be connected Hardware Independence to one port, a daisy wheel to another, and a laser printer to a PC/LPD uses the FTP Software third, up to the capacity of the PC. PC/TCP kernel as its transport and PC/LPD can handle requests from any the kernel is included with PC/LPD. host with an Lpr client (or Because PC/LPD uses the PC/TCP equivalent) which is Berkeley 4 BSD kernel, it will operate on Ethernet compliant. This lets networks with a TM, Starlan, and Token Ring variety of operating systems use a networks. Over 40 different network single print server. interface cards are supported, making it possible to use practically any interface card on Operation the market. PC/LPD will run on IBM PCs, ATs, PS/2s, or compatibles, so PC/LPD provides multiple printer a site can often install PC/LPD service and runs continuously as a without buying any additional daemon taking over the PC. In other hardware. words, the computer can't be used for any other applications while PC/LPD is running. PC/LPD on a PC LANWatch(R) now Supports Ethernet runs roughly as fast as similar Packet Drivers programs run on a DEC(R) MicroVAX(R), since the printer is by James VanBokkelen the speed limiting factor. A new version of LANWatch which Keyboard commands are available to uses the Packet Driver specification operate PC/LPD. You can control for monitoring Ethernet and Starlan individual printers, control each networks is now in beta test. We printer's queue, and control overall expect quite a lot of sales from the server operation. A variety of new LW-110, because it has the operating data can be displayed, potential to open up a large number including statistics on printer of network interfaces which operation and network connections, previously could not be used by printer queue lists, and debugging LANWatch. There are a few hurdles information. that need to be dealt with first, however. PC/LPD is compliant with the line printer daemon as documented in the As of version 2.04, the Generic Berkeley UNIX 4 BSD programming Ethernet PC/TCP uses only a minimal documents. It accepts lpr, lpq, and the "basic" level of the functions lprm commands from other hosts. defined in the Packet Driver These programs are available in a specification. The following wide range of software which functions (from the "basic" level) implements the Berkeley utilities, are required: including Berkeley UNIX, MultiNet for VMS and PC/TCP. However, you driver_info() Get Packet Driver should note that some UNIX characteristics implementations (and other operating systems) do not have these utilities get_address() Return MAC address available. You should check to make of the interface sure your software has an _____________________________________________________________________________ June 1990 4 access_type() Request set_rcv_mode() to mode 6 fails, demultiplexing of a LANWatch can't continue. packet type Second, the Packet Driver LANWatch release_type() End demultiplexing calls access_type() only once, where of a packet type PC/TCP calls it five times. For its one "handle", LANWatch passes a send_pkt() Transmit a packet, typelen argument of 0, "match all blocking until done packet types", in order to see all the different protocols on the wire. In addition, the 2.04 kernel will If the Packet Driver doesn't support make use of the following "high this, LAN-Watch exits with an error performance" level function, if message, giving the error code available: returned by the driver. get_parameters() Read extended The Ethernet drivers in the next driver configuration release (Ver-sion 6) of the Clarkson info freeware Packet Driver collection will most likely support LANWatch; This is used primarily to the alpha versions we've seen so far determine if the Packet Driver have run perfectly. We are also issues an interrupt which the kernel enhancing our own NDIS to Packet can hook to do further processing Driver converter module to support after the driver has dismissed a LANWatch (although it will only be hardware interrupt. If this is useful where the NDIS drivers available, the kernel can provide themselves support "match all better transport layer performance addresses"). We hope that other for our TCP/IP NetBIOS product. driver developers and maintainers will seriously consider adding The Packet Driver LANWatch uses "extended" functionality over the the same functions that 2.04 does, next year or so; LANWatch is one except for send_pkt() and incentive, another will be our get_parameters(). It also breaks planned addition of IP Multicast new ground, in two ways. support to the PC/TCP kernel, which will require either the First, it uses two "extended" set/get_multicast_list() functions, Packet Driver functions: or support for mode 5 (all multicast packets) in set_rcv_mode(). set_rcv_mode() Change interface's address matching INT 0x14 Telnet get_statistics() Read interface hardware counters by Dave Burdelski set_rcv_mode() allows the PC/TCP's INT 0x14 BIOS emulator, application to control what sort of tnglass, uses the definition of INT hardware address matching the 0x14 that is published in the IBM interface performs. Various levels AT(R) Hardware Technical Reference. are specified, but LANWatch requests This manual presents the BIOS mode 6, "receive all packets". If listings and specifications. All the Packet Driver indicates that it four "basic" functions of INT 0x14 doesn't support "extended" functions are implemented. These functions, (in the "functionality" value which are invoked based on the value returned by the driver_info() found in the AH register at the time function), or if the call to _____________________________________________________________________________ June 1990 5 of the INT 0x14 instruction, are /* c is character from INT 0x14 listed below. service routine. */ register byte c; Function AH AL { /* i.e. if bits 0 and 1 are set: */ initialize port 0 initializa- if ((c&3)==0x03) { tion byte /* then request TELNET binary mode! */ send a character 1 character to tn_rqst_bin(tn); be sent } else { /* i.e. if bit 1 is set and bit 0 is receive a character 2 character not: */ read returned if ((c&0x02)==2) { here /* tell host WONT BINARY (i.e. 7 bit mode) */ status 3 status value tn_bin_off(tn); returned } } On return (IRET) from the INT 0x14 } routine for function AH=3, a status word is returned in AX. This functionality has been tested extensively in the 2.04 patch level Our current current implementation 1 implementation of the tnglass does not implement the so-called program. One way to verify that the "extended" BIOS functions (AH > 3) code really works is to write a as published in the PS/2(R) simple test program through the PC's Technical Reference Manual or the debug program. Let's assume that FOSSIL extensions. you have connected to a host called "random" using the tnglass emulator An ambiguity that seems to be with command.com (the DOS command plaguing emulator writers is that interpreter), as follows: before you can expect to send, or (especially) receive, 8-bit data on c:> tnglass.exe host -e \command.com the connection, you must specify that the word length is 8 bits. Now, enter debug by typing "debug" This initialization is done through at the prompt. Then, assemble the the AH == 0 function of INT 0x14. following program through debug, According to the BIOS listing in the using the "a" command at the debug IBM AT Hardware Technical Reference prompt. Line by line, this program Manual, the two low order bits of AL does the following: determine the word length: mov ah,0 ; sets up for the if (bit 1) == 1 and (bit 0) == 0, ; initialization of the then the word length is 7 bits ; RS232 port. (Subfunction ; initialize) if (bit 1) == 1 and (bit 0) == 1, then the word length is 8 bits mov al,3 ; Sets bits 0 and 1 ; indicating an 8-bit word The snippet of code below (from ; length the tnglass program) shows what our software does when asked for various int 14 ; Call the BIOS INT 0x14 word lengths. ; (Serviced by PC/TCP). int14_bin(c) nop ; hang out here! _____________________________________________________________________________ June 1990 6 Execute this down to the "nop" Boulder, CO 80301 instruction, by using the debug "g" (303) 447-9251 command. Watch your LAN analyzer to get a packet trace of the Grafpoint TM makes emulators for transaction between you and the the Tektronix 4010, 4014, 4105, remote host, and you should see 4107, 4109, and 4115 terminals, as several packets traded. The PC well as the VT52 and VT100 terminals should send TCP Telnet from DEC. For more information, contact: data "ff fd" (Telnet IAC followed by Telnet "DO BINARY") and the Grafpoint remote host, if it understands 4340 Stevens Creek Boulevard Telnet binary mode, should reply San Jose, CA 95129 with "ff fb" (Telnet IAC followed by (408) 249-7951 Telnet "WILL BINARY"). Next, you should see the PC send "ff fb" TEEMTALK TM from Pericom emulates (Telnet "WILL BINARY") and the the Tektronix 4105, 4111, and 4207 remote host should respond "ff fd" terminals, as well as the VT52, (Telnet "DO BINARY"). If these VT100, VT220, and VT241 ReGIS packets are traded as specified terminals from DEC. For more here, the connection is now binary, information, contact: and 8-bit data will flow correctly. Pericom This is an empirical test to 2291 205th Street, Suite 103 determine that the code works as Torrence, CA 90501 stated. Until a standard has been (213) 618-9190 established, we hope that this will serve as a guide to INT 0x14 SmarTerm(R) 240 from Persoft(R) emulator implementors who wish to emulates the VT52, VT100, VT220, interoperate with the PC/TCP BIOS VT240, VT241 with ReGIS graphics, emulator. and VT340 with 16-color ReGIS graphics terminals, as well as the 2392 terminal from Hewlett-Packard. Editor's Note For more information, contact: A number of terminal emulation Persoft, Inc. software packages have been written 465 Science Drive to use INT 0x14 and the PC/TCP Madison, WI 53711 kernel to communicate with a remote (519) 886-5320 host over a network. Some are listed below. Additional products The Reflection(R) series of will probably work with PC/TCP, if products from Walker, Richer & Quinn they normally communicate through emulates the 700/92, 2392A, 2623A, the COM ports and can be 2624B, and 2627A color graphics reconfigured to use INT 0x14. terminals; Tektronix 4014 terminal; and the VT52, VT102, VT220, and DCS makes emulators for the VT241 with 16-color ReGIS graphics Tektronix(R) 4014 and 4105 terminals from DEC. For more terminals, as well as the VT(R)100 information, contact: and VT220 terminals from DEC. For more information, contact: Walker, Richer, & Quinn 2825 Eastlake Avenue East Diversified Computer Systems Seattle, WA 98102 3775 Iris Avenue, Suite 1B (800) 872-2829 _____________________________________________________________________________ June 1990 7 The INTEROP 89 Network: From one of from their booths for the INTEROP its builders show to use for emergency shortages and changes in topology, some just by Stev Knowles hours before the doors opened Wednesday at noon. The network ran Reprinted with permission from smoothly, once the show opened. The ConneXions -- The Interoperability only complaint that made its way Report TM, published monthly by back to us was one conference Interop, Inc., 480 San Antonio Road, attendee who felt that everything Suite 100, Mountain View, CA 94040. was running too smoothly. There was Phone 415-941-3399. no one running around, typing hurriedly at a terminal, while 5 people stood around him with Introduction concerned expressions on their faces. Whoever you were: sorry There are infinite reasons to go about that! to the INTEROP trade show. Daniel Dern went through several in his ConneXions article in the November The Real Heroes 1989 issue. In the past, when I have gone to INTEROP, I have done a There are two people who deserve wide variety of things, going to most of the credit for the success class sessions to learn about new of the INTEROP show network, Peter protocols my company had an interest de Vries, and Lisa Robertson, both in, attended BOF sessions concerning from Interop. These are the people issues from the extremely boring to who spent months planning out the the excitingly fascinating. I had physical topology, adding up the hoped that INTEROP 89, the biggest lengths of cable runs, locating fan- INTEROP held so far, would be a out boxes and concentrators, place to renew friendships with the coordinating with the vendors who people I rarely see, and to talk loaned all the hardware that makes a (all right, argue) about current hot network run, making sure that topics. Unfortunately, I didn't get everyone was listed in all the much of a chance to do any of these appropriate lists for their network things. Fortunately, what I did get drops, and making sure that the 25 to do was infinitely more or 30 cases of assorted soda that interesting than arguing about the technical people consumed were routing with people who already have always cold. This is an important their minds made up. point: if they remembered to keep the soda cold, you can assume that Before I get lost in detail, allow they had thought of everything else me to clarify some points. I am too. They had. sure that there will be people whom I will not mention here, who deserve You may wonder why I am writing some credit for their assistance in this, if all the credit for this INTEROP 89. The number of companies goes to Peter and Lisa. Well, to be and individuals who provided honest, so do I. I was invited to hardware and assistance is join about 10 other people to put monumental. Rarely have I seen so together, play with, and take apart many companies help each other out, an arbitrarily complex network. A with loans of equipment and lot of people take this very technical expertise to bring seriously, and ignore that this is everyone's networks up with a in reality a challenging goal. minimum amount of hassle and stress. During the brief time it is up, it Several companies provided equipment _____________________________________________________________________________ June 1990 8 is a pretty amazing toy. The fact understood why we were there, and it was kept running in a near are the people who put in 18 hour flawless condition was due to our days from Friday until Wednesday pride in its magnificence. Our afternoon. Some of the other people grand toy should appear nothing less who should be mentioned are Geoff than awe-inspiring to other people. Baehr (Sun), Jeff Burgan (NASA As for our motivation for putting Ames), Ronnie Hueter, Phil Almquist forth all this effort, if you have (Stanford University) and Chris to ask, you wouldn't understand the Lynch (Interop). answer. This article is not intended to Network drops give one a coherent story about what went on in creating the show-net for Friday morning was spent marking the INTEROP 89 show, but rather to the approximate locations of the provide highlights of the network drops in the booths (some difficulties we ran into, and kind souls had already marked the explanations of the options as we booth locations for us). We just saw them, and our decisions and did a rough estimate of the reasoning behind the decisions. It locations; since they would be is possible (actually, probable) hanging from the ceiling some 30 that any given decision could have feet above us, we could afford a bit been made better, but one must of leeway in the exact locations realize that the people who produced that the wires had to be dropped to. the shownet did so in 4 days, Peter had spent some serious time working 18 hour days. involved in network planning; when we arrived there were maps of everything we could possibly want. An early start There were maps for all the major media types, UTP (Unshielded Twisted Friday morning started early; 5:30 Pair), Thin Ethernet, Thick am wake up call, 6:00 am meeting. Ethernet, 802.5, and FDDI fiber. Valerie Collins (Interop) had reserved a room for us every morning These maps were further sub- at 6:00 am for breakfast until divided into "work sheets." Work Wednesday. The first breakfast sheets were parts of the master maps included about 20 people, including in smaller, workgroup sized reps from some of the vendors who allotments. The long, "backbone" were loaning us hardware and some of thick Ethernet runs were on one map, the more esoteric other toys we the thin Ethernets fanning out from needed (fiber optic cable, breakout each fan-out location were on boxes, line testing equipment). others. Within 3 hours it became clear Media like UTP and thick Ethernet that there were about 10 serious which had a large number of drops people here. They were, in no were put on maps based on sections particular order, Peter, Lisa, Steve of the show floor (east, middle, and Larbig (Interop), Dave Bridgham west). Copies of these maps were (Epilogue Technology), Karl Auerbach used by crews of people who went off (Epilogue), John Romkey (Epilogue), to do their assigned tasks. When Eric Brunner (Tule Network they were done, and when time Services), Alan Brunner permitted, someone else would go (consultant), Shelly de Vries over the work they had done to make (Interop), and myself (FTP sure everything was in place. As it Software). These are the people who _____________________________________________________________________________ June 1990 9 turned out, we still ended up them back to the ceiling in the missing about 3 network drops. hall. The benefit to this Consid-ering there were several approach was that it would be hundred, this, I suppose, should not quick to implement. The reflect badly on us. problems were many, as we saw it. Since we were running cable Our greatest resource was having through a door, it would have to some people on tap who have been be left open all day and night, doing this for a long time, and requiring another security guard along with being up on all the to sit by it. We decided that latest technology, are pretty while we could put a connector flexible people, willing to scrap in the cable at this point, and the well laid foundation Peter gave disconnect it at night, that us and strike out on our own when it would be cutting off e-mail appeared that our previous plans service after show hours. We were not going to work. also decided we could not use Fortunately, Peter was as agile as one of the main doors, since we were, and changes were quickly people would tend to pull on the hashed out and implemented. wires and such. We also decided it was not aesthetically pleasing. Running wires out from the exhibit hall 2) We could punch a small hole through the cinder block wall. We found, as we were running This was eventually decided to wires, that we had been given be antisocial (!) incomplete information about the building we were working in. We had 3) The convention center includes been told that one could get into floor access plates for power the ceiling and thereby cross from and phone all over the show the exhibit hall to the main floor. These open into the concourse running along the front of garage under the convention the building. This we anticipated center. The people we sent back using to provide cable runs to the out were to determine how to get mail centers (one each on the east the cables from the garage to and west ends of the convention the drop locations along the center), the microwave relay (for concourse in the front of the the mail center in the Fairmont building. Unfortunately, there Hotel, across the street), and a set was no access between the garage of information terminals being and the concourse that didn't provided by Prime for the concourse involve propping open a door. area. We found that the exhibit We decided it would be worse to hall's cinder block wall went to the leave these wires hanging out so roof, and was sealed to the roof far from us, and we didn't want with mortar to provide fire walls. to leave a guard down with them. We sent off people to detail the building, and to provide 4) The final suggestion, and the recommendations for how to get the eventual solution, involved cable out of the exhibit hall and careful study of how a large into the concourse area in light of building is put together, and this obstacle. We were brought back playing on the weaknesses that several suggestions: show up. The convention center is not really one building, it 1) We could just run the wires out is actually 4 buildings built 5 one of the doors, and then run _____________________________________________________________________________ June 1990 10 inches apart. The space between hotel. The problem was getting from them is covered by aluminum the room the receiving dish was in panels. This allows the to the room the mail center was in. building to expand and contract The hotel had told us they would be as it heats and cools. As it able to provide a connection between turns out, you can also get from these rooms. We assumed they meant the exhibit hall to the they would be able to connect us concourse through these through the phone drops in the two passages. The problem here is rooms by jumpering them together in that these expansion joints tend a phone closet. They thought all we to travel under walls. needed to do was run modems between the two rooms, and they had refitted the rooms to have modular connectors Expansion joints in them. We decided that we had benefited from investing the time in The convention center breaks the sending someone to detail the large hall up into 4 pieces, each convention center, so we sent slightly larger than the preceding someone over to look at the hotel. one. This is why the concourse in Their suggestion we went along with. front of the building is much larger It involved running a UTP line from on the east side than the west side, the microwave room, to the since the walls of the exhibit hall stairwell, down to the floor the move farther towards the front of terminal room was on, and over. the building as you go the the west. Fortunately, both rooms were on the The building joints tend to appear same side of the hotel. This at the locations where the walls jut involved us using additional UTP out to make the next hall bigger. transceivers, which we had over- ordered, but required an additional We eventually managed to pass UTP fan-out unit, since you cannot cables through the expansion joints, connect UTP transceivers "back to under the walls, and into the back." Discussions with the concourse area. This involved an SynOptics engineers resulted in them incredible amount of contortion, but removing a unit from one of their seemed the best solution to the demo racks and handing it over to us problem. We could then continue the to use elsewhere. run under the floor (in the expansion joint), to the front of The final problem with the mail the building, where we planned to centers, that resulted in us opening have the mail center runs, and the them late, after the INTEROP microwave and Prime terminals were tutorials had started, was a set up. We ended up spending a lot shipping delay in the fiber we of time trying to figure this one expected to run along the front of out, and it involved some careful, the buildings. We decided to wait real-time planning to allow us to for the fiber to show up because our remove expansion joint covers when shipping contacts told us it was in the convention center was not being transit, and because we already had used by INTEROP or other tenants the appropriate hardware on hand to (this usually happened after allow us to use it. After Monday, midnight or 1 am.). we decided that we were going to assume the fiber was never going to show up, and huddled together for Meanwhile, at the Fairmont hotel another solution. The final decision was based solely on what we At this point, we had a problem had left over. We ended up running with the microwave connection to the _____________________________________________________________________________ June 1990 11 thin Ethernet from the drop to both stranded connectors, and solid wire. ends of the building for the mail The runner reported that the store centers. This introduced yet he had bought the connectors at only another problem. We had no had that one type. One of the equipment on hand to bridge from the people who had extensive experience thick Ethernet to the two thin showed us a connector he had in his Ethernets we were running. A quick pocket. It was much easier to use, talk to the engineers at the since it required no special tools. Wellfleet booth produced a box we It had a flared back, where one could use as a bridge for the could lay the wire strands in duration of the show. This was channels, and screwing the cap on taken from the demo machines they caused teeth to bite into the wires. had brought for use in their own booth. This proved to be much quicker than the crimp-on connectors (even after we had gotten some crimping Connectors tools), and much more reliable, but had the disadvantage that the part One serious problem we had was that stuck out of the wall was so with the UTP connectors. These were wide that the fan-out boxes we used standard eight conductor telco could not have connections in every modular phone type connectors. jack; we had to leave empty jacks Initially, the shipment of between them. While Peter had connectors we got could not be planned extra jacks, he had not located. We sent out a runner to planned on twice as many jacks. get more. One thing that people Fortunately, Synoptics was able to putting on something like this provide us with additional fan-out forget is the value of someone to go boxes for the duration of the show. out and get the things you forget. The more you can keep your crews Here we got bitten by several doing the stuff they do best, the problems. The connectors did not more progress you make. Having arrive on time, and the replacements someone who is willing to spend the that were bought were the incorrect day running around and getting type (even the packaging did not parts, and lunch, for the crews is imply that there were two types). very important, and may be one of Fortunately, someone was able to put the most generally useful people you us in touch with a vendor who would will have around. The runner extend us immediate credit for the returned with crimp-on connectors. connectors we needed and let us go Since I have seen these connectors and pick them up from them with no before, I gave the crew starting on notice. Some of the crimping tools it a quick class in crimping the we had did not work correctly, and connectors with a pair of channel- there were not enough of them if lock pliers. They went off to be they had. The module we were using productive. to test the lines was defective, since the adapter we plugged the cable into had a broken wire. Some Two types of these things we could have taken care of before hand, like more Later they returned proclaiming a crimping tools, and testing the startling lack of success. As more tester out on a known good wire. people got involved, we discovered Some of it, like the two types of that there are two types of crimp-on crimp-on connectors, we should have connectors; one for stranded wire, known about, but most of us had one for solid. We, of course, had _____________________________________________________________________________ June 1990 12 never run into these problems had a "phone" drop, but not the before. thick Ethernet drop they expected. Very quickly we started putting up UTP or "All I got was a phone signs at the entrance doors telling connection!" people that the UTP drops above their booths were indeed network One of the big problems was caused drops, and not to worry about it. by people not paying enough We surrounded the INTEROP booth with attention. We ran a lot of UTP. these signs also. This did not seem UTP was used for several reasons: to stem the tide very much. When people would come to ask a question, * It was cheap, and easily we would just point to the closest available. sign, and usually they would go away. Sometimes they would ask * It was easy to work with, and anyway. We made the mistake of (we thought!) easy to telling them they would have "thick" connectorize. Ethernet or "thin" Ethernet on the forms they filled out. * It didn't weigh much, so we could connect more runs to one What we should have done was ask anchor point on the ceiling. if they wanted a drop for a transceiver cable or a T-connector. * Using off-the-shelf technology, Since we were vague about this, some we could provide either thick people did not bring enough of their or thin Ethernet access from own equipment. Fortunately, a lot it. of companies brought more than they thought they could possibly use, and The topology was simple, and were very helpful in loaning involved the use of several fan-out connectors and cables to each other. boxes. Fan-out boxes can be Seeing competitors helping each cascaded to a certain degree other fix problems is unusual in (cascading is hanging one fan-out today's competitive market. It is a box under another, instead of shame that the customers that came attaching them all at the same to see the shownet never got to see level). The destination end can this wonderful cooperation. either provide the user with a "thick" connection (really, a place to plug in a transceiver cable), or Routing a "thin" connection (a place to attach a T-connector). Actually, there were few machines Unfortunately, fewer people than we on the INTEROP show floor that would have thought had heard of this misbehaved on the community wires. technology. This would not be a Some booths had routers between problem, but people were being told their networks and the shownet that we had finished running wire internetwork. When possible, we got around the show floor (they were the network addresses they were concerned since their "roady cases" using internally, and allowed them were being moved in, filling the to be RIPed around the shownet aisles, making it impossible for us floor. Since the shownet was to move cherry-pickers around the connected to the outside world by booths. People were constantly several routes (including NASA Ames coming up to us telling us that they and the NSFNET NOC in Ann Arbor), we were very sensitive to routing _____________________________________________________________________________ June 1990 13 issues. We allowed only one of the several key people. People who know shownet routers to peer with routers what you are trying to do very well, outside of our internetwork, and we and are not so rigid that they only allowed it to advertise a route cannot accept suggestions from out to the show-net class B network. We of their scope of knowledge. were not interested in the dynamic Maximizing communications is routing on the NSFNET backbone important. Interop provided the finding out that there was a low wiring crew with 8 walkie-talkies. cost route from the NSFNET NOC to These went far in speeding up the NASA Ames across the shownet. decision making process. We spent considerably less time tracking While this in itself would have people down to collect their opinion been an impressive advertisement for about crises and planning changes the adaptability of TCP/IP to expand than we would have had to otherwise. and contract its routes based on It allowed us to propagate changes availability, we decided that it to the crews faster, and to quickly would be a hard thing to convey to reassign people to fight crises that the people visiting INTEROP 89, and were always springing up. Since we it would also add to the load of had several people around who, while what we thought would be an over- not being knowledgeable about what crowded backbone. we were trying to do, were willing to learn how to do the jobs we As it turns out, the shownet needed done, willing to follow backbone was very crowded. The orders, willing to go run out and network monitoring programs we had pick up supplies, or call people to available were scrolling packets by track down equipment, we freed other faster than one could read. people to do more complex jobs. Fortunately, any of the routers we put on the shownet backbone could be Keeping people working on what configured to broadcast routing they do best greatly increases the information only on certain subnets, amount of work a crew can and accept information only from accomplish. Since the people certain hosts. assembled here all knew each other, we could quickly put together a good This way, we did not have to worry group to evaluate change proposals, about someone's mis-configured BSD- or make suggestions about how to fix derived system trying to peer with the problems we encountered. Often the shownet backbone routers. Since we found that even when incomplete there were several "fan-out" boxes groups met to make decisions, the in use, short term network outages groups would make a decision that were restricted to small physical the missing people would agree with. parts of the much larger logical Some people find it hard to run a network, when one thin drop from a large project by consensus. We fan-out box went unterminated, found it to be a process that usually because someone was adding speeded us up rather than slowed us another segment and T-connector to down. Having several people with a it, the other drops in the fan-out complete "world view," who could be box did not notice the outage, and counted on to give an authoritative continued to function as before. answer on 95% of the questions asked, allowed people trying to get work done to move faster since they Lessons learned could get their approval from a closer source. What makes an endeavor like this possible? The cooperation among _____________________________________________________________________________ June 1990 14 Conclusion officers of the corporation, and the staff. An asterisk next to the name All in all, this was a pleasurable of a board member indicates a experience. I got to work closely previous employee who continues to with some of the best people in our maintain close ties with the business, and I was able to learn a company. lot from the people around me. We ended up with something that the trade show attendees were excited Members of the Board about, and impressed with. I found the vendors in our industry to be Nancy Connor made up of smart, helpful people Robert Goodnow willing to give their competitors a John Romkey * hand when they forgot equipment or Mark Rosenstein * cables needed to connect equipment James VanBokkelen to the shownet. While it was a very draining experience, I would be glad to do it again. We have already Officers started thinking about INTEROP 90. From what I hear, since people Nancy Connor - Chairman of the thought it went so smoothly this Board time, we are going to have to make James VanBokkelen - President it more complex and flashy next year Stev Knowles - Vice President, to impress the attendees. I can't Engineering wait to start playing with it! Roger Greene - Vice President, Sales & Marketing STEV KNOWLES is a hacker at FTP Robert Goodnow - Treasurer Software. (This referring to the Nancy Connor - Secretary "Old School" of hackers, when being a hacker was a title bestowed upon you by others. It is not the only Technical Support title bestowed on Mr. Knowles by others, but it is the only one we Dave Burdelski - Manager could get past the publisher.) Mr. Paula Burke Knowles had many diverse words (some Vicka Corey of encouragement and some not) for Jennifer DeSouza the INTEROP 88 shownet engineering Margaret Forsythe team, and therefore was drafted for Peter Grasso the 89 team. He is currently involved in the IETF Benchmarking Methodology Working Group, and Documentation watches over several projects at FTP. He has no degrees, Caroline Lange - Manager certificates, or warrants Patricia Hanagan outstanding for his arrest. He can Dave Shute be reached by Internet e-mail as stev@ftp.com. Development The FTP Software Staff Bill Rust - Director Glen Daniels Since FTP Software has been Joel Gartland growing so fast over the past year, Michael Greenberg we would like to introduce you to Sue Gustafson the members of the board, the _____________________________________________________________________________ June 1990 15 Karen Kohn Tammy D'Amore - Receptionist Bill Lampman Kelly Lewis - Receptionist Benjamin Levy Marguerite Walsh - Administrative Gary Malkin Assistant Mark Mason Rehmi Post Paul Selkirk Walter Stickle Sales Bob MacFadgen - Manager Joe L'Italien Russ McDonnell Colin Mahan Bill Shaw Jim Worley Nancy Zipf Penny Zirkle Marketing Bruce Campbell - Director, Product Marketing Raghu Nath - Channels Marketing Manager Cris Shuldiner - Product Manager Production Byron Robyn - Manager Glen Barry Robert Casa Kevin Gosson Keith Nardone Finance Jane Reid Larissa King Sharon Smith Riki Weiner Dana Mercer Administration Janet Marcisak - System Administrator Frances Selkirk - Technical Assistant