LAN Technology Scorecard

Compiled by
John Wobus,
LAN info

This is a table (i.e., a "scorecard") of various technologies for implementing Local Area Networks on various media. For explanation of the acronyms, many of which are mine, see the keys below.


The Table

Technology Rate      Dist  Packet   Wiring         As Of Standard     Products
---------- --------- ----  -------  ------------   ----- ------------ ---------
4Mb Token  4Mbps     570m  TokenR4  Type1/Type2    10/94 IEEEdone     Out
4Mb Token  4Mbps     260m  TokenR4  C3/C4/C5       10/94 IEEEdone     Out
4Mb Token  4Mbps     ?     TokenR4  Fiber          10/94 IEEEdone     Out
16Mb Token 16Mbps    ?     TokenR16 Type1/Type2    10/94 IEEEdone     Out
16Mb Token 16Mbps    ?     TokenR16 Fiber          10/94 IEEEdone     Out
16Mb Token 16Mbps    160m  TokenR16 C3/C4/C5       10/94 IEEEdone     Out
64Mb Token 64Mbps    ?     TokenR16 ?              2/95  Rumor        ?
100Mb Tkn  100Mbps   100m  TokenR?  C5             9/97  IEEEdiscuss  ?
100Mb Tkn  100Mbps   2km   TokenR?  Multimode      9/97  IEEEdiscuss  ?
1Gb Token  1Gbps     ?     TokenR?  Fiber/TP       9/97  IEEEdiscuss  ?
LocalTalk  230.4kbps 300m  AppleTk  C1             12/94 Proprietary  Out
ARCNET     2.5Mbps   100f  ARCNET   RG62           7/94  ANSIdone     Out
TCNS       100Mbps   100m  ARCNET   RG62A/U        7/94  Proprietary  Out
TCNS       100Mbps   150m  ARCNET   Type1          7/94  Proprietary  Out
TCNS       100Mbps   900m  ARCNET   Fiber          7/94  Proprietary  Out
TCNS       100Mbps   100m  ARCNET   4PC5           7/94  Proprietary  ?
ThinWire   10Mbps    185m  Ethern   ThinWire       12/93 IEEEdone     Out
ThickWire  10Mbps    500m  Ethern   ThickWire      12/93 IEEEdone     Out
10BASE-T   10Mbps    100m  Ethern   2PC3           11/93 IEEEdone     Out
FOIRL      10Mbps    1km   Ethern   Multimode      12/93 IEEEdone     Out
10BASE-FL  10Mbps    2km   Ethern   Multimode      2/94  IEEEdone     Out
10BASE-FB  10Mbps    2km   Ethern   Multimode      2/94  IEEEdone     Out
10BASE-FP  10Mbps    500m  Ethern   Multimode      2/94  IEEEdone     ?
PMD        100Mbps   2km   FDDI     Multimode      11/93 ANSIdone     Out
oldCDDI    100Mbps   100m  FDDI     UTP            9/94  Propriet/Obs Out
oldTP-FDDI 100Mbps   100m  FDDI     2PC5/Type1/2   9/94  Obsel        Out
SDDI       100Mbps   100m  FDDI     STP            12/95 Propriet/Obs Out
SMF-PMD    100Mbps   40km  FDDI     Singlemode     2/94  ANSIdone     Out
LCF-PMD    100Mbps   500m  FDDI     Multimode      12/95 ANSIdone     Out
TP-PMD     100Mbps   100m  FDDI     2PC5/Type1/2   12/95 ANSIdone     Out
FDDI-II    100Mbps   40km  FDDI     Sing/Mult/etc  12/95 ANSIalmost   Out
old4T+     100Mbps   100m  Ethern   4PC3/4/5       11/93 IEEEstarting Late 93/94
old100B-X  100Mbps   100m  Ethern   2PC5/Type1/Fib 1/94  IEEEsubmtd   Out
100BASE-TX 100Mbps   100m  Ethern   2PC5/Type1     7/95  IEEEdone     Out
100BASE-FX 100Mbps   412m  Ethern   Multimode      7/95  IEEEdone     Out
100BASE-T4 100Mbps   100m  Ethern   4PC3/4/5       7/95  IEEEdone     Mid 95
100BASE-T8 ?         ?     Ethern   8PC?           9/94  ?            ?
100BASE-T2 100Mbps   100m  Ethern   2PC3/4/5       7/95  In works     Yrs away
100BASE-SX 100Mbps   412m  Ethern   Multimode      11/97 ?
FD100BS-SX 100Mbps   670m  Ethern   Multimode      11/97 ?
100VG-AnyL 100Mbps   100m  Eth/To   4PC3/4/5       2/96  IEEEdone     Out
100VG-AnyL 100Mbps   150m  Eth/To   2PC5/Type1     8/95  IEEEprop     ?
100VG-AnyL 100Mbps   2km   Eth/To   Sing/Mult      6/95  IEEEratified Late 94
100VG-AnyL 100MbpsFD ?     Eth/To   ?              8/95  IEEEprop     ?
100VG-AnyL 400Mbps   ?     Eth/To   C5             8/95  IEEEprop     ?
1G-AnyLAN  960Mbps   2km   Eth/To   Fiber          1/96  IEEEdiscuss  97
100VG-AnyL 4Gbps     ?     Eth/To   Fiber          8/95  IEEEprop     ?
T100       50Mbps    ?     Ethern   2PC3           6/94  Proprietary  Out
T100       100Mbps   ?     Ethern   4PC3           6/94  Proprietary  Out
WaveBus    100Mbps   2km   Ethern   Multimode      12/96 Proprietary  Out
WaveBus    100Mbps   500m  Ethern   Multimode      12/96 Proprietary  Out
WaveBus    100Mbps   10km  Ethern   Singlemode     12/96 Proprietary  Out
SwEthern   10Mbps    100m  Ethern   2PC3           11/93 No change    Out
PACE       10Mbps    100m  Ethern   2PC3           11/94 Proprietary  95Q1
SwFastE    100Mbps   100m  Ethern   2PC5/Type1     9/95  No change    Out
FDSE10     10MbpsFD  100m  Ethern   2PC3           3/96  IEEEballot   ?
FDSE10     10MbpsFD  2km   Ethern   Multimode      3/96  IEEEballot   ?
FDSEold    10MbpsFD  100m  Ethern   2PC3           3/96  To be obsel  Out
FDSE100    100MbpsFD 100m  Ethern   2PC5/Type1     3/96  IEEEballot   ?
FDSE100    100MbpsFD 2km   Ethern   Multimode      3/96  IEEEballot   ?
IS-LAN     16Mbps    100m  Ethern   2PC3           2/96  IEEEratified Out
1000BASE-X 1Gbps     ?     Ethern   Fiber/C?       6/96  IEEEdraft    Late 97
1000BS-TX  1Gbps     100m  Ethernx  4PC5           6/98  IEEEdiscuss  ?
1000BS-CX  1Gbps     25m   Ethernx  TW             3/97  IEEEdraft    ?
1000BS-SX  1Gbps     100m  Ethernx  Multimode      3/97  IEEEdraft    ?
1000BS-LX  1Gbps     550m  Ethernx  Multimode50    3/97  IEEEdraft    ?
1000BS-LX  1Gbps     300m  Ethernx  Multimode62.5  3/97  IEEEdraft    ?
1000BS-LX  1Gbps     3km   Ethernx  Singlemode     3/97  IEEEdraft    ?
10GENET    10Gbps    ?     Ethern?  ?              3/99  Discussion   ?
CoENET     .1-1Gbps  100m  Ethern   2PC5/Type1     5/96  Proprietary  12/96
TUTE       100Mbps   1500f Ethern   1PC3           12/95 Proprietary  Out
SwToken    16Mbps    300m  TokenR16 Type1/C4       7/94  No change    Out
FDToken    16MbpsFD  300m  TokenR16 Type1/C4       2/94  ?            Mid 94
SwFDDI     100Mbps   2km   FDDI     Multimode      12/93 No change    Out
FFDT       100MbpsFD 2km   FDDI     Multimode      11/94 Proprietary  Out
FDFDDI     100MbpsFD 2km   FDDI     Multimode      11/94 Proprietary  Out
FDDIFOL    2.4Gbps   ?     ?        Fiber          10/95 Inactive     Future?
HIPPI-PH32 800Mbps   25m   HIPPI-FP 100Pair        8/94  ANSIdone     Out
HIPPI-PH64 1.6Gbps   25m   HIPPI-FP 100Pair        8/94  ANSIdone     Out
HIPPI-Ser  800Mbps   10km  HIPPI-FP Singlemode     3/95  Speced       Out
FibreChan  100Mbps   1.5km FibreChn Mult/Cx/TP     8/94  ANSIdone     Out?
FibreChan  200Mbps   10km  FibreChn Sing/Mult/Cx/TP8/94  ANSIdone     Out
FibreChan  400Mbps   10km  FibreChn Sing/Mult/Cx   8/94  ANSIdone     Out?
FibreChan  800Mbps   10km  FibreChn Sing/Mult/Cx   8/94  ANSIdone     Out?
FibreChan  1600Mbps  ?     FibreChn ?              2/95  Speced       ?   
FibreChan  3200Mbps  ?     FibreChn ?              2/95  Speced       ?    
FC-EP      4Gbps?    10km? FibreChn Sing/Mult/Cx   8/94  ANSIproposed Late 94
Myrinet    640Mbps   25m   Myrinet  Copper         8/94  Proposed     Out
STS3cUNI   155Mbps   ?     Cell     Sing/Mult      12/93 AFpublished  Out
DS3UNI     45Mbps    ?     Cell     ?              12/93 AFpublished  Out
E3UNI      ?         ?     Cell     ?              4/94  ?            Out
100MbpsUNI 100Mbps   ?     Cell     Multimode      12/93 AFpublished  Out
155MbpsUNI 155Mbps   2km   Cell     Multimode      12/93 AFpublished  ?
155MbpsUNI 155Mbps   100m  Cell     Type1/Type2    12/93 AFpublished  ?
155MbpsUNI 155Mbps   ?     Cell     C5             4/94  AFapproved   ?
TAXI/140   140Mbps   ?     Cell     Multimode      11/94 Proprietary  Out
SONET/OC1  51Mbps    ?     Cell     Sing/Mult      11/93 Done         ?
SONET/OC12 622Mbps   ?     Cell     Sing/Copper    4/94  AFproposed   ?
SONET/OC48 2.4Gbps   ?     Cell     ?              11/93 Done?        ?
SONT/OC192 10Gbps    ?     Cell     ?              9/96  ANSI?        ?
ATMT1      1.5Mbps   2Km   Cell     UTP            10/94 Proprietary  Out
ForumT1    1.5Mbps   2Km   Cell     UTP            10/94 AFapproved   Out
LST1       1.5Mbps   2Km   Cell     UTP            10/94 Proprietary  Out
E1         2Mbps     ?     Cell     ?              ?     ?            Out
J2         6Mbps     ?     Cell     ?              3/95  AFadopted    ?
ATM25      25Mbps    100m  Cell     C3/4/5/Type1   2/95  AFselected   Out
CIF        10Mbps    100m  Cell     2PC3           5/96  AFproposed   ?
STS-1/2    25Mbps    ?     Cell     C3             2/95  AFrejected   ?
51MbpsUNI  51Mbps    100m  Cell     C3/4/5/STP     8/94  AFballot     ?
TUT51      51Mbps    ?     Cell     C3             1/95  AFproposed   95H1
TUT155     155Mbps   65m   Cell     C3             8/95  Proprietary  Out
CAP64      155Mbps   ?     Cell     C3             1/95  AFproposed   ?
ATMFrm100? 100Mbps   ?     Cell     Copper         2/94  AFproposed   ?
Fire100    100Mbps   4.5m  ?        FireSTP        5/97  IEEEdone     Out
Fire200    200Mbps   ?     ?        FireSTP        10/96 IEEEdone     ?
Fire400    400Mbps   ?     ?        FireSTP        10/96 IEEEdone     ?
P1394.2    800Mbps+  ?     ?        FireSTP        10/96 IEEEsubm     ?
P1394a     800Mbps+  ?     ?        FireSTP        10/96 IEEEsubm     ?
Fire1600   1.6Gbps   ?     ?        FireSTP        5/97  ?            ?
SerialE    ?Gbps     ?     ?        ?              4/96  IEEE?        ?
USB        12Mbps    ?     ?        ?              8/96  ?            ?
WideBand   800Mbps   100m  Ethern   4PC5           1/98  Proprietary  Out
WideBand2  3.2Gbps   ?     Ethern   ?              1/98  Proprietery  Now
---------- --------- ----  -------  ------------   ----- ------------ ---------
Technology Rate      Dist  Packet   Wiring         As Of Standard     Products

Key to Chart Headings

Some don't have short names so I had to improvise.
The raw rate: note that the raw rate of the transmission media is often not the bottleneck, and in at least one case (HIPPI) represents a maximum possible raw rate rather than a fixed rate.
Maximum distance from a hub to a user station through the given technology. In the case of the various Ethernet technologies, it does not include transceiver cable (MAU) distance. Sometimes there are other limits, e.g. the distance limitations imposed by CSMA/CD.
what type of packets it transmits. Important in determining whether bridging between this technology and others needs either encapsulation or translation. Kinds: Cell, Ethernet, FDDI, TokenRing Cell refers to ATM cells.
Type of wire or fiber supported.
"As Of"
the date of the source of the latest information incorporated for the line. For example, if someone sends me e-mail during 12/94 saying that a standard was done in 7/94, then I put 12/94 in this field.
current state of the standard
current state of products

Technology Key

4Mb Token
4 Mbps Token Ring: IEEE 802.5 4 Mbps.
16Mb Token
16 Mbps Token Ring: IEEE 802.5 16 Mbps. Distances between hub and station actually depend upon equipment and ring configuration. One opinion: C3 is really not suitable, 90m maximum distance is safest for TP, 160m has been seen to work on C5, and for more 160m, fiber is best.
64Mb Token
64 Mbps Token Ring: Nothing more than a rumor at this point. If IBM has considered it, clearly they would consider foregoing it simply because of the number of alternatives in the works.
100Mb Tkn
100 Mbps Token Ring: proposal from Tolly Round Table and IEEE 802.5 Interim Meeting on High Speed Token Ring , Aug. 26-27, 1997, at which time an agressive schedule is being proposed for a copper version based on the 100BASE-TX physical standard, that will complete standardization during Summer 1998. Efforts toward a fiber version based on 100BASE-FX will follow. Packet size not yet set but likely to be same as 16Mb Token.
1Gb Token
Efforts toward a version based on the Gigabit Ethernet standards are planned after the initial 100Mb Tkn stanard.
Uses low-grade twisted pair (originally coax) as daisy chain, bus or star. TCP/IP commonly encapsulated in AppleTalk packets over LocalTalk.
ANSI 878.1. Developed by Datapoint. Uses Token Bus access method. Aside from RG62, also runs on twisted pair through baluns. The 100ft limit is with the use of ARCNET Passive hubs. With an active hub, a 2000ft run is possible, and I've heard 400ft 2PC3 quoted also.
Proprietary LAN technology developed by Thomas-Conrad as a drop-in replacement for ARCNET offering 100Mbps. Also uses Token Bus access method.
ThinWire Ethernet or IEEE 802.3 10BASE2.
"ThickWire" Ethernet or IEEE 802.3 10BASE5.
IEEE 802.3 10BASE-T.
IEEE 802.3 FOIRL: "Fiber Optic Inter-Repeater Link". Note that the standard is a bit vague on whether this can be used to connect a station to a repeater.
IEEE 802.3 10BASE-FL: replacement for FOIRL.
IEEE 802.3 10BASE-FB: part of the new IEEE 802.3 10BASE-F spec: "Synchronous Ethernet" which is a special-purpose link for linking repeaters that allows the limit on segments & repeaters to be enlarged. Note that it was designed for inter-repeater links though it can be used to link to end stations and there are applications where this can be of use.
IEEE 802.3 10BASE-FP: part of the new IEEE 802.3 10BASE-F spec: support for a passive star configuration.
FDDI "Physical Lay Medium Dependent" part. When "PMD" is used by itself, it usually refers to the usual kind of FDDI physical layer that uses multimode fiber. Note that FDDI also uses it as a more generic term, referring to different FDDI PMD's. Operates at 125Mhz using a 4B/5B NRZI encoding.
What was formerly known as CDDI: Proprietary technology developed by Crescendo. FDDI-like technology adapted to unshielded twisted pair. Efforts to standardize it have yielded TP-PMD. Cisco, which bought out Crescendo dropped the trademark on 'CDDI', blessing its use as a common industry term for TP-PMD.
"Twisted Pair-FDDI": Technology to run FDDI over twisted pair proposed by 11 companies that have now joined in with TP-PMD.
"Shielded Distributed Data Interface". Proposal to run FDDI over shielded twisted pair by IBM and 10 or so other FDDI companies. Effectively superseded by the ANSI standards now.
FDDI "Single-Mode Fiber" PMD. Runs further than PMD. Published as ANSI X3.184-1993 Rev. (4.2, 12/7/92).
FDDI "Low-Cost Fiber" PMD. Less expensive than PMD. Published as X3.237-1995, Rev. 2.2.
FDDI "Twisted Pair Physical Layer Medium". Also often called CDDI. ANSI specification for FDDI-like service over UTP or STP(Type1 or Type2). Being standardized by ANSI X3T12. Is now at Rev 2.2 (T12/95-022, 1 March 1995) and published as ANSI X3.263.1995. Operates at 125Mhz using a 4B/5B encoding, but uses MLT-3 encoding (a three-state encoding) instead of old CDDI & PMD's NRZI encoding to reduce emissions. The latest changes to TP-PMD were for a "base line wander problem". Some "TP-PMD" products sold do not incorporate the latest revisions the specification.
FDDI II. I've also seen it called isoFDDI. This is not necessarily an FDDI follow-on and doesn't change the media & technology so much as the way it is used. It standardizes a convention for setting aside bandwidth for various applications, by supporting multiple 6.144Mbps channels. Note that it can use all the PMD's for FDDI, thus for example can run 100m on 2PC5. A number of standards comprise FDDI-II (Hybrid Ring Control, Enhanced Media Access Control, Enhanceed Physical Layer Protocol, Enhanced SMT Common Services, Enhanced SMT packet Services, Enhanced SMT Isochronous Services).
4T+: old name for 100BASE-T4. It may be that some prestandard equipment was released under the name 4T+.
100BASE-X: old name for 100BASE-TX & 100BASE-FX which now would refer to pre-standard equipment.
100Mbps CSMA/CD 802.3/Ethernet-like LAN also known as Fast Ethernet. There are three flavors: 100BASE-TX, 100BASE-FX, and 100BASE-T4 (see individual descriptions below). Repeaters or hubs would be necessary to adapt and there will be a media-independent interface which NICs can support so that an external adaptor determines which flavor can be attached. Packets are identical to 802.3 packets (with bit-times 1/10 the time), but the nature of CSMA/CD requires that the overall radius of the net be limited to 1/10 the size of 10Mbps Ethernet. A typical maximal system would be hubs on a very short backbone (up to 5 meters), the (repeating) hubs supporting links up to 100 meters. Single-hub networks allow up to 325 meters (e.g. 225 meters of fiber for one link and 100 meters of twisted pair for any other link). Extension of the net beyond this would require a switch, router, or bridge. Fiber links employing the CSMA/CD but with no hubs can run 450 meters, and full-duplex links (i.e., with CSMA/CD "disabled") can run 2km. It is obvious that without (at minimum) switches, that this technology will be limited to connecting a few offices to a server at most. All three have been defined by IEEE 802.3 in the IEEE P802.3u/D2 Supplement.
One of the three flavors of 100BASE-T. Basically a renaming of the twisted-pair variant of 100BASE-X. Borrows the physical characteristics of FDDI's TP-PMD, but uses Ethernet framing & CSMA/CD. A Media Independent Interface will allow a single interface card to use either this or the other flavors of 100BASE-T, just as the AUI allows a 10Mbps 802.3 controller to use any of its several types. See also 100BASE-T above. Also in development is a method for NIC and hub to negotiate between 10BASE-T and 100BASE-T (called Nway) and I'm guessing this refers to 100BASE-TX. The method consists of extensions to the 10BASE-T link integrity test pulse sequence with pulses that signal 100BASE-T capability. This should allow "dual capability" NIC cards to be installed before a network is brought up to 100Mbps speed.
One of the three flavers of 100BASE-T. Basically a renaming of the fiber variant of 100BASE-X. Borrows the physical characteristics of FDDI's normal fiber PMD, but uses Ethernet framing & CSMA/CD. Note that the physical media supports distances up to 2km, but in a CSMA/CD environment (i.e., any 802.3 except full-duplex links), there is a maximum limit of 412 meters between on the size of the entire network. A Media Independent Interface will allow a single interface card to use either this or the other flavors of 100BASE-T, just as the AUI allows a 10Mbps 802.3 controller to use any of its several types. See also 100BASE-T above.
One of the three flavers of 100BASE-T Basically a renaming of 4T+. Uses 8B6T (three-state: 8 bits encoded into 6 trits) encoding and 25MHZ clocking, and in addition to the two pairs traditionally used in the manner of 10BASE-T, also has two pair used in bidirectional half-duplex fashion. Among other things, this means that this particular kind of Ethernet cannot be made full duplex without the use of more pair. A Media Independent Interface will allow a single interface card to use either this or the other flavors of 100BASE-T, just as the AUI allows a 10Mbps 802.3 controller to use any of its several types. I've read on the net that 100BASE-T4 cannot be used on 4 pairs within a 25 pair C3 cable because there is too much near end crosstalk, but that it does work in 4-pair cables (i.e. using all the pairs). See also 100BASE-T above.
Something I heard mention of once: possibly mistaken or defunct.
New initiative in June 1995. Being worked on by IEEE 802.3y.
As of 11/97, I heard of the development of fiber Ethernet operating at 850nm instead of 100BASE-TX's 1300nm. It uses 4B/5B encoding. The advantage of such a scheme is that 100BASE-FL operates at 850nm, so 10/100 autosensing could be developed.
"Full Duplex 100BASE-SX". Should be able to reach 100BASE-SX's non-CSMA/CD physical distance limitation of 670 meters.
"100VG-AnyLAN" (VG means "Voice Grade"): Originally a proposal to IEEE 802.3 for a 100Mbps Ethernet-like network, later relegated to IEEE 802.12. Formerly known as 100BASE-VG. Uses Demand Priority media access method and when using 4 pair, Quartet Signalling, which operates at 30Mhz, using a 5B/6B encoding which keeps emissions low by using only relatively balanced ratios of the two states. The spec will indicate that the 4-pair version will operate at 100m on C3 or 150m on C5. Under good conditions, 200m on C3 and 350m on C5 have been accomplished. My impression is that this is over 4 pair using Quartet signalling. The 2 pair versions of it operate at 120MBaud, using the 5B/6B encoding to provide 100Mbps. I've read on the net that unlike 100BASE-T4, the 4-pair version of 100VG-AnyLAN will work on 4 C3 pairs within a 25-pair cable. I've seen the net radius quoted at 600m for C3 and 122m for C5. I read that the IEEE "ratified" 100VG-AnyLAN on June 13, 1995. As of August, 1995, plans have been announced for full duplex 100VG-AnyLAN as well as 2PCat 5, 400Mbps, and 4Gbps versions.
1 Gigabit variation of 100VG-AnyLAN being developed by the same IEEE committe (802.12). At the time I heard of this, not all was decided, including the exact speed at which it will run: I've heard mention of 500Mbps, and have also heard that it would borrow the physical layer of 800Mbps Fibre Channel. Word as of early 1998, is that work on this is suspended.
Proprietary high-speed Ethernet scheme that runs 50Mbps on 2 pair C3 or 100Mbps on 4 pair C3. Developed & sold exclusively by LAN Performance Labs.
Proprietary high-speed Ethernet scheme that runs 100Mbps on multimode or Single Mode fiber.
Switched Ethernet: really the same as Ethernet as far as standards go; just the use of Ethernet to attach a switch (i.e., multiport bridge) to a client or smaller group of clients. Typically, 10BASE-T would be used to interconnect the switch and the client. Since "Switch" has become an industry buzzword, things that used to be called a "Bridge" are now known as switches, especially models with more than two ports. Current industry jargon uses the term "Store and Forward Switching" for what used to be called bridging as opposed to "Cut Through Switching" and vendors of the two types will argue that one of these methods is superior to the other. Cut Through Switching means sending the packet on before it is completely received and implies very low-latency which is good for network services that alternate packets in each direction, but it transmits runts and erroneous packets. A modification of cut-through switching delays transmission for an Ethernet slot time and avoids transmitting runts. The latency of a Store and Forward switch depends upon the model--it is something to check on. They do not transmit either runts or erroneous packets and for doing so, they adhere to the IEEE 802 standard for bridges. The typical cut-through switch is clearly harder to build than a store-and-forward switch and was done that way on purpose to increase performance, but many customers feel a store and forward switch with good latency offers no significant performance disadvantage. Other factors aside from sheer performance: spanning-tree algorithm for loop detection; flow-control; Number of MAC addresses supported per port.
Priority Access Control Enabled: 3Com's proprietary technology designed to reduce Ethernet overhead and latency by making an Ethernet switch and computer "take turns" sending packets, probably equivalent to a 2-node token bus. In effect, it takes advantage of the fact that with switched Ethernet, each "true Ethernet" has just two nodes on it which can cooperate to help utilization & latency. Plans are to adapt it to 100Mbps Ethernet also. Requires both special switch and changes to the user's computer.
Switched Fast Ethernet: can have physical characteristics of any of the 100BASE-T or 100BASE-F variants.
10Mbps Full Duplex Switched Ethernet (FDSE): standard under consideration by IEEE 802.3x; a variant of Switched Ethernet which does not use CSMA/CD, but uses slightly-modified network interface cards to send & receive packets simultaneously. Presumably based on 10BASE-T for most clients, and cannot be based on ThinWire or ThickWire Ethernet. Since the distance limitations imposed by CSMA/CD are eliminated, the only problem is how far a line can be driven. The standard will be independent of speed so it will apply to selected versions of 100Mbps and future faster versions of 802.3. There have been proposals to support up to 50km on single mode fiber.
pre-standard Full Duplex Switched Ethernet (FDSE): while there are vendors who offer full duplex switched Ethernet now, none include all the features of the standard in progress (IEEE 802.3x; see FDSE10), specifically its flow-control features.
100Mbps Full Duplex Switched Ethernet (FDSE): part of the IEEE 802.3x standard. 100BASE-T4 will not be able to support this while 100BASE-FX and 100BASE-TX will, given suitable electronics. 100BASE-TX remains electrically limited to 100 meters while 100BASE-FX can run 2km. See also FDSE10.
IEEE 802.9A, often called Isochronous Ethernet or isoEthernet, the latter which is a National Semiconductor Trademark). An adaptation of 10BASE-T to support another 6Mbps of possibly multiplexed synchronous data along with the 10Mbps CSMA/CD. Proposed by National Semiconductor. Uses 4B/5B encoding instead of 1B/2B encoding to carry the larger amount of data. The 6Mbps is divided into ISDN channels (i.e. 96B+D+E; E for Ethernet) which do not carry Ethernet-style packets but can carry a data in the formats supported by ISDN channels. I've also heard of a 96kbps "M for maintenance" channel. The 6Mbps channel (group of channels) is compatible with ITU H.320, a specification used for videoconferencing that can use multiple ISDN-stle B channels. Promoted by incAlliance.
Gigabit Ethernet. This is being discussed by IEEE standards-making groups. Both the IEEE 802.3 and IEEE 802.12 committees are interested in the concept and IEEE 802.3z has been formed to standardize such a thing. Target is a standard approved by March 1998 with prestandard products coming available the first half of 1997. Things being discussed include a CSMA/CD version as well as a full-duplex version; and a physical layer that consists of a modified Fibre Channel interface (1250 MBaud signalling for 1Gbps data rate). Stated objectives are 100m for horizontal cabling, 550 meters for backbone cabling, and 3km for campus cabling. See separate entries below for the likely physical standards (1000BS-TX, 1000BS-CX, 1000BS-SX, 1000BS-LX). CSMA/CD imposes a relation between the minimum length of packet, the speed of the network and the maximum size of the network and at 1Gbps, the kinds of packets used on 10Mbps and 100Mbps Ethernet would restrict the size to 50 meters. Plans are to effectively increase the minimum frame size by holding carrier up after the smallest packets, thus allowing CSMA/CD to be used up to 200 meters. Another feature is senders' option to catenate small packets together while sending. See IEEE 802.3z and IEEE 802.3ab. The Gigabit Ethernet Alliance has been formed to facilitate its standarization. Also, see
1000BASE-TX: Gigabit Ethernet over Catagory 5 Twisted Pair. See 1000BASE-X. A separate IEEE committee, 802.3ab, is working out this standard. The original target for completion was mid to late 1998. Goal is to support 100 meters over 4 pair of Category 5 UTP, but it is possible that a finer degree of standardization of cabling will be necessary, i.e. it may be that not all Category 5 UTP will be able to handle it. It uses four pair with a clock rate of 125Mhertz, using 5 level signaling, to carry 2 bits each clock cycle with some room left over for forward error correction & clock synchronization. See IEEE 802.3ab.
1000BASE-SX. Gigabit Ethernet short-wavelength laser over 62.50 or 50 micron multimode fiber. See 1000BASE-X. Uses Fibrechannel parts. Goal is to assure 100 meters minimum. Tests show 300m on 62.5 micron fiber and 550m on 50 micron fiber. See IEEE 802.3z.
1000BASE-LX. Gigabit Ethernet long-wavelength laser over single modeo multimode fiber. See 1000BASE-X. The long light wavelength is 1300 nanometers: equipment to use this is more expensive but can support longer distances. Goal is to assure 550 meters on multimode fiber or 3 kilometers on single mode fiber. Tests show 850m is possible over mutimode. See IEEE 802.3z.
1000BASE-CX. Gigabit Ethernet short copper connection. See 1000BASE-X. Goal is to provide interconnections within a room, e.g. between switches. Goal is 25 meters over Balanced, Shielded Media ("TW Type Cable"). See IEEE 802.3z.
As yet unnamed 10 Gigabit Ethernet. As of 3/97, reports are that there are informal talking among the usual participants of the IEEE 802.3 work groups.
a proprietary scheme from Cogent to reverse-multiplex No name yet, but one assumes it will be 1000BASE-something. a large stream packets over 1-10 Fast Ethernet links, giving 100Mbps to 1Gbps throughput.
Tut Systems Inc's Long Distance Ethernet Repeater spec. Uses propritary noise-rejection methods to run Ethernet over telephone lines at distances to 1500f.
analog of Switched Ethernet: each client gets a separate ring that interconnects it with a high-speed packet switch.
IBM scheme to add switching to token-ring hubs that would allow full-duplex linking to individual computers using modified token-ring adaptors. Has the same wiring characteristics as token ring.
Switched FDDI: really the same as FDDI as far as standards go: acts like a very fast multiport FDDI bridge. Basically the DEC GIGAswitch. Note that it can use all the PMD's for FDDI, thus for example can run 100m on 2PC5 or 40km on SMF.
FDDI Full Duplex Technology: DEC's proprietary modified FDDI that runs full duplex instead of "token passing" on their GIGAswitch FDDI switch and adaptors. They have a patent on their own method of detecting automatically whether a link is full-duplex, which requires extensions to SMT, the FDDI ring-management protocol, which they license to other vendors. Note that it can use all the PMD's for FDDI, thus for example can run on 100m on 2PC5 or 40km on SMF. DEC Licenses the technology to multiple companies.
(for Full Duplex FDDI: I don't know the true name) I heard once that Cabletron is planning to offer full duplex FDDI but know nothing about any details or cooperation between vendors. Note that it can use all the PMD's for FDDI, thus for example can run 100m on 2PC5 or 40k on SMF.
FDDI Follow On. An idea floated in the ANSI X3T9.5 committee for a method of multiplexing FDDI and ATM data on fiber at speeds up to 2.4Gbps.
ANSI HIPPI with a 32-bit-wide data transfer. Standardized by ANSI X3T11.
ANSI HIPPI with a 64-bit-wide data transfer. Standardized by ANSI X3T11.
HIPPI-Ser "HIPPI-Serial"
HIPPI over fiber or coax; either as a transparent fiber extender for HIPPI-PH32 or HIPPI-PH64, or as a native host interface. Being standardized by ANSI X3T11.
Fibre Channel: an ANSI standard for high-speed data transfer over fiber designed to do what HIPPI can do and more. It can be made to emulate HIPPI as well as various disk buses (SCSI, IPI, Block Mux) and can also carry LAN protocols, (IP, etc). Happens to use the same 8B/10B encoding as some flavors of ATM. Being standardized by ANSI X3T11 as ANSI X3.230.199x. Note: I've seen the numbers 266Mbps and 1.062Gbps quoted so I don't know how those fit into the 100/200/400/800/1600Mbps heirarchy of speeds which I heard about previously. Also, I've seen the Baud Rates quoted as 133, 266, 531, and 1062 MBaud.
Fibre Channel Enhanced Physical Interface: ANSI proposal for a faster version of Fibre Channel: 4Gbps or 16Gbps.
Developed by Myricom. Full-duplex 640Mbps channels connecting hosts and switches. Uses 0.4" shielded, multiconductor cable (type CL2(?)).
ATM Forum SONET STS-3c UNI, 155.52Mbps. Also called OC3. When you see references to 155Mbps ATM running over fiber, it almost always refers to this though the UNI spec includes another (see 155MbpsUNI, below).
ATM Forum DS3 UNI, 44.236Mbps, a standard for carrying ATM cells on a DS3 telecommunications line. It borrows the Physical Layer Convergence Protocol from IEEE 802.6. Not actually a LAN protocol, but rather for WANs.
ATM (Forum?) E3 UNI, 34Mbps, a standard for carrying ATM cells on an E3 telecommunications line. E3 is the European equivalent to North American DS3. Unlike the DS3 UNI, the E3 UNI does not borrow the PLCP from 802.6. Not actually a LAN protocol, but for WANs.
ATM Forum 100Mbps multimode fiber private UNI. Often called TAXI. FORE developed this, borrowing optical characteristics & basic encoding of FDDI, using AMD's "TAXI" chips. When you see the phrase "TAXI" or "100Mbps ATM", it almost always means this.
ATM Forum 155Mbps private UNI. In two flavors: multimode and shielded twisted-pair. The multimode version is incomplatible with STS3cUNI. This version is for private networks only and presumably will be less expensive. The multimode fiber version uses the same 8B/10B encoding as FibreChannel. I heard that a C5 version has been proposed. When you see the phrase "fiber 155Mbps ATM", it almost always doesn't mean this, but rather the SONET STS-3c version. When you see references to 155Mbps ATM over twisted pair, it almost always means this.
ATM server/switch specification developed by FORE and often called 140Mbps TAXI. Like 100MbpsUNI, adapted from FDDI. The only difference between this and the 100MbpsUNI is that this has the clock cranked up 40%. FORE no longer manufactures it since the standard for the similar STS3cUNI has come out, but they support existing devices.
We refer to a SONET-based ATM variant using fiber.
ATM Forum SONET/SDH UNI specification with OC3c/STM-1 framing. Variants of the underlying SONET spec use different media: multimode, single mode (short reach or long reach).
We refer to a SONET-based ATM variant using fiber.
We refer to a SONET-based ATM variant using fiber.
We refer to a SONET-based ATM variant using fiber.
I don't know the actual name for a T1-based method supported by some vendors. It borrows the Physical Layer Convergence Protocol from IEEE 802.6. E1 is also presumably supported similarly. Not actually a LAN protocol, but for WANs.
I don't know the actual name for a T1-based ATM Forum UNI. Rather than DS1, this uses HEC framing as specified in ITU-T recommendation G804. Not actually a LAN protocol, but for WANs.
I don't know the actual name for a proprietary method of supporting ATM through T1 which is sold by Lightstream. Not actually a LAN protocol but for WANs.
I don't know the actual name for an E1-based UNI. "E1" is a telecom term for a 2Mbps T1-like telecom standard used in Europe. Here we refer to a method of running ATM over it.
T2-like 6.3Mbps ATM interface. T2 (between T1 and T3) is not used in the US, but a version of it is widely used in Japan. Not actually a LAN protocol, but for WANs.
25.6Mbps private UNI proposed to ATM Forum by IBM. Borrows physical characteristics of 16Mb Token Ring (both operate at 32Mhz), gaining extra capacity by using FDDI's 4B5B encoding (on top of an NRZI code) in place of Token Ring's differential Manchester encoding. As of February 1995, it had been selected by the ATM Forum's PHY Group over STS-1/2. Note that AVIDIA Systems and Efficient Networks have decided to market this under the name "25Base-T", which AVIDIA has tradmarked, in order to stress that it will run over the same cable as 10BASE-T.
"Cells in Frames", also called ATM Emulation. A specification for putting ATM Cells in Ethernet frames. I listed the characteristics of 10BASE-T, but it could be run over any type of 10Mbps Ethernet as well as Token Ring. Cornell University is the primary developer. See
25.9Mbps private UNI proposed by PMC-Sierra, an ATM chip supplier (I've also seen the term UniPhy-25 applied to it). It is proposed as an alternative to the IBM proposal. It uses SONET framing, thus an interface can share more hardware the SONET-based versions of ATM. Based on the forum's fractional SONET framing technology. As of February 1995 it was rejected by the ATM Forum's PHY working group.
ATM Forum's "Mid-range Physical Layer Specification for Category 3 Unshielded Twisted-Pair". Uses AT&T's 16-CAP (Carrierless Amplitude Modulateion, Phase Modulation; a 16 constellation modem-type modulation scheme) line coding to transmit the signal. The transmission convergence layer (framing) conforms to the STS-1 SONET standard. It can run 160m on C5 cable. Also as a option, the equipment can be made to support longer distances by dropping to 1/2 or 1/4 speed. 1/2 speed uses CAP-4 encoding and 1/4 speed uses CAP-2 encoding.
I don't know the actual name for Tut Systems' reported proprietary 51Mbps over C3 ATM technology. Uses NRZ encoding.
I don't know the actual name for Tut Systems' reported proprietary 155Mbps over C3 ATM technology. Also uses NRZ encoding. I've read statements that seem to suggest this is supposed to be compatible with the C5 version of the 155MbpsUNI (or actually, an enlargment of that spec). The problems with running 155Mbps over C3 include single attenuation and near-end crosstalk. The TUT155 technology uses Next cancellators to keep crosstalk to a minimum.
CAP-64 based 155.52Mbps Physical Media Dependent layer proposed to the ATM Forum. CAP-64 (stands for Carrierless Amplitude Modulation/Phase Modulation with a 64-point constellation) is analogous to CAP-16 (see 51MbpsUnNI above) etc. It achieves its speed despite running the cable at no more than 30Mhz and can comply with FCC Class-A and Class-B radiation requirements.
I don't know the actual name. ATM Forum UNI for 100Mbps over some sort of copper cable. I believe it is just 100MbpsUNI making use of FDDI's TP-PMD rather than the older fiber PMD.
100Mbps "FireWire". FireWire is a popular name for the data interconnection defined as IEEE 1394-1995. It is a serial bus more or less designed to interconnect computers and I/O devices. Among the features: isochronous data transfer (which is supposed to allow multiplexing data with bandwidth guarantees), up to 63 devices on a bus, hot pluggable devices (typical for LANs, but not for many types of I/O interconnects). It uses a special cable made up of two sheilded twisted pairs and power which is used to allow daisychaining through devices that are powered off, and possibly to provide power some low-power devices. Signal is sent on one STP as NRZ, with the other being a strobe signal which transitions only in cases where two successive bits are the same. There is interest among vendors of consumer and professional electronics in the fields of video, audio, and games. Though it is really conceived as an improvement on peripheral busses like SCSI, other types of LANs were also started with such ideas (e.g. FDDI), and it seems that something like this could become the dominant technology of the "one-room Home LAN". An effort is also underway to standardize a FireWire bridge (IEEE P1394.1). Also, the IEC has approved it as its standard 1883.
200Mbps FireWire/IEEE 1394-1995. See Fire100.
400Mbps FireWire/IEEE 1394-1995. See Fire100.
proposed 800Mbps+ version of FireWire/IEEE 1394. that is not backward-compatible with Fire100, Fire200, and Fire400.
proposed 800Mbps+ version of FireWire/IEEE 1394 that is backward-compatible with Fire100, Fire200, and Fire400.
"Serial Express". Serial Express is Intel's name for the data interconnection defined as IEEE 1394.2. Supports higher speeds and longer distances than IEEE 1394 (FireWire).
USB "Universal Serial Bus"
designed to be an improved technology for the kind of communication accomplished with a microcomputer's serial port. Runs at 12Mbps, supports up to 127 devices through daisy chaining; supports connection and disconnection while the computers are powered up.
Proprietary technology from WideBand Corporation (note that the name of the technology is similar to a common communications term). Raw throughput is 666.7Mbps downstream (which is divided into two channels) and 333.3Mbps upstream which is reduced by 20% through use of 8B/10B coding. Communication is full duplex, using Ethernet-compatible packets. See
Future faster version of WideBand.

Packet Types Key

An AppleTalk packet. 5-603 bytes.
An ARCNET packet. 1-508 bytes (excluding 254-256).
An ATM 53-byte cell. Note: there are various proposals for how typical packets will be broken into cells and restored.
An Ethernet packet: 64-1518 bytes.
In the case of Gigabit Ethernet, I heard a rumor that the minimum packet length would be increased.
Ethernet or Token Ring style packet.
An FDDI packet: 0-4478 bytes of data (total of 20 to 4495 bytes including header).
A Fibre Channel packet. 128-2112 bytes.
Not really a packet: the framing specific to HIPPI.
A Myrinet packet. Myricom's software allows 8,368 bytes, the hardware allows much longer.
A Token Ring packet. Allows longer packets than Ethernet, among other things. 4-megabit Token Ring allows 4500 byte packets, 16-megabit token ring allows 17800 byte packets.

Wiring Key

? Pairs (used in designators 2PC3, 4PC3, 2PC5, 1PC3, etc.; for example, 2PC3 means "2 Pair Catagory 3 Unshielded Twisted Pair")
HIPPI 100pair cable.
Coax or Cx
some sort of coax: don't know which kind
some sort of copper connection
Category 1 Unshielded Twisted Pair
Category 3 Unshielded Twisted Pair
Category 4 Unshielded Twisted Pair
Category 5 Unshielded Twisted Pair
Special FireWire/IEEE 1394 Shielded Twisted-pair cable. It includes power.
Multimode fiber.
Multimode fiber with a 50 micron core.
Multimode fiber with a 62.5 micron core.
Screened Twisted Pair: not listed yet, but I might someday hear about which technologies that can use it
Shielded Twisted Pair
Ethernet/IEEE 802.3 Normal "Thick" Coax.
Ethernet/IEEE 802.3 ThinWire Coax.
Balanced, Shielded Media, TW Type Cable (twinaxial cable or twinax: similar to coaxial cable but with two internal conductors).
IBM Type 1 STP.
IBM Type 2 STP.
Unshielded Twisted Pair

Vendor support/products

(Note: this includes vendors' plans and announced interests as well as their current products)

4Mb Token
IBM, etc
16Mb Token
IBM, etc
64Mb Token
100Mb Tkn
1Gb Token
Apple, many vendors
Thomas-Conrad, Contemporary Control Systems
DEC, etc
DEC, Intel, Xerox, etc
many vendors
many vendors
NCR, many vendors
Chipcom, IBM
many vendors
IBM, 3Com, Madge, Network Peripherals, SysKonnect
SynOptics, National Semiconductor, DEC, Cisco, NPI, 3Com, SysKonnect, UB Networks, ODS, Cabletron, IBM, Team Advanced Systems, Alfa, Chipcom, Distributed Systems International, Gambit, Proteon, Interphase, Memorex, Network Peripherals, NetWorth, Raylan, Rockwell, Xyplex, Xylan, GEC Plessey (chip), Motorola (MC68840 chip), Wolfson (chip)
Loral Federal Systems, Distributed Systems International, AWA Defense Industries (Austrialia)
Grand Junction
SynOptics, Intel, Accton, Sun, DEC (DECchip 21140), 3Com, National Semiconductor, ODS, Cisco, Thomas-Conrad, Plexcom, TI, Compaq, Grand Junction, Fujitsu, SMC (SMC91C100 chip), Amber, Asante, Seeq (84C300 chip), LANNET, NetWorth, Apple, Netcom, Cogent, Asante, Bay, ANT, Olicom, Proteon, Dayna, Farallon, Rockwell, Kalpana, Interphase, HP, Shomiti, D-Link, Acacia, Adaptec, GigaLabs
Grand Junction, National Semiconductor, Sun, SynOptics, David, Intel, DEC, 3Com, Cabletron, Wellfleet, Chipcom, Racal-Datacom, SMC, NCR, GEC Plessey (chip), NetWorth, NBase, Farallon, Cnet, Wolfson (chip) (See list for Fast Ethernet Alliance), Lite-On, IBM, Sonic, RNS, Dayna, PlainTree, Cisco, Transition, XLNT, Linksys, HyNEX, Xircom
SynOptics, Plexcom, Grand Junction, NBase, Focus, PlainTree, Cisco, Cabletron, Transition
3Com, ATT, DEC, SynOptics, Intel, NCR, NetWorth, Broadcom (BCM5000 chip) (See list for Fast Ethernet Alliance)
HP, ATT (Regatta 100 chip set), IBM, Proteon, UB Networks, SMC, ODS, DEC, D-Link, Andrew, Racore, Racal InterLan, Thomas-Conrad, 3Com, Alfa, TI (chips), Compaq, Cisco, D-Link, Ragula, Newbridge, Compex, Katron, Madge, Wellfleet, Bay, MultiMedia, Plaintree, Chipcom, Motorola (chips), AMCC (chips), Pericom (chips), PureData, Kalpana, Interphase
ATT, Compac, HP, Motorola, TI
LAN Performance Labs
Kalpana, Artel, Alantec, Grand Junction, LANNET, Cabletron, 3Com, SynOptics, Synernetics, Hughes, Calios, SMC, NBase, NetWiz, IBM, Xedia, HP, Matrox, Plaintree, Chipcom, Amber, Network Peripherals, Retix, NiceCom (bought by 3Com), Fibronics, Fibermux, Onet, Agile, Ascom Timeplex, Bytex, OST, Plexcom, Bay, UB, Xylan, NetWorth, XNET, CrossComm, Allied Telesys, Cisco, NetVantage, ODS, Lantronix, Whitetree, Xpoint, XNET, Xedia, ANT, Klever, ORNET(ONET), XLNT, GigaLabs, HyNEX
NBase, Cabletron, Cisco, FORE, Galileo (GT-48002 chip)
Cabletron, Kalpana, IBM, 3Com, Compaq, National Semiconductor, NCR, SEEQ, Texas Instruments, Cogent, HP, ODS, Sun, SynOptics, NBase, NetWiz, DEC, Hughes, LANNET, Alantec, Grand Junction, AMD, GigaLabs
SynOptics, Intel, Kalpana, Grand Junction, 3Com, XLNT
National Semiconductor, Ascom-Timeplex, Apple, IBM, ATT, Ericsson, Microsoft, Pacific Bell, Siemens/Rolm, Zydacron, 3Com, Novell, Sun, Dell, Silicon Graphics, Oracle, Networks AB, Luxcom, Incite/Intecom, Ascom-Nexion, Luxom, Quicknet, Dialogic, Future Labs, ITT, Incite, MCI, Primary Rate, Teleos, VCON, (See list for incAlliance)
Sun, 3Com, Compaq, Granite, Amdahl, Cisco, Packet Engines, NBase, Network Peripherals, Cabletron, Rapid City, Extreme Networks, Mammoth Networks, Prominet, Alteon, Bay, MMC Networks, UB, XLNT Designs, FORE, PlainTree, Ancor, GigaLabs, Acacia, AMCC (chips), Scalable, Xylan, Foundry (was StarRidge), Adaptec, Essential Communications, XaQti, Acclaim, Berkeley Networks, Digital, HP, IBM, Intel, Ipsilon, Madge, Neo, YAGO, G2 Networks, Foundry, Netcom, Network Appliance, Silicon Graphics, (See list for Gigabit Ethernet Alliance)
IBM, NetWiz, Ace North Hills, Madge, Chipcom, Centillion Networks, Bytex, ODS, SMC, SynOptics, Kalpana, 3Com, NetEdge, Bay, Cisco, NetVantage, Nashoba, Xylan, Connectware, Olicom, Northern Telecom, UB
DEC, Centillion Networks, XLNT
DEC, Distributed Systems International, CMD, CIsco
Acri, AMP, Ampex, AMCC, Avaika, Broadband Communications Products, Chi Systems, CNT, Convex, Cray Computer, Cray Research, Datatape, DEC, E-Mass, E-Systems, Essential Communications, Fujitsu, Genroco, GES, HP, Honeywell, Hytech, IBM, Intel, Lockheed, Loral Defense, MasPar, Maximum Strategy, Meiko Scientific, Methode, Myriad Logic, NEC, NetStar, NSC, Pacific Title Digital, PsiTech, Silicon Graphics, Siemens, Sony, Sun, Tera, Texas Instruments, Texas Memory Systems, Thinking Machines, Triplex Systems, TRW, Vertex, Zitel
Cray, Network Systems, Broadband Communications Products, PsiTech
Avaika, Broadband Communications Products, Essential Communications, NetStar, Tera
Ancor, HP, IBM, Sun, Western Digital, GENROCO, Emulux, VLSI, AMP (see FCSI), Interphase, Seagate, Augment, GigaLabs, Jaycor, Finisar, Gadzoox, Seagate.
ATM (general)
FORE, Newbridge, GTE, Fujitsu, ATT, Alcatel, General DataComm, Hughes, LightStream, NEC, NET, Network Systems, Northern Telecom, ODS, StrataCom, SynOptics, Telematics, TRW, ADC Kentrox, Cabletron, Cascade, Cisco, DEC, FastComm, Interphase, NetEdge, Efficient, ZeitNet, First Virtual, Agile, Whitetree, PMC-Sierra (PM7345 chip), Connectware, Thomas-Conrad, ATML, RADCOM, Tricord, Astarte, Philips, IBM, Mikroelektronik Anwendungszentrum Hamburg, Brooktree, National Semiconductor, TI, TransSwitch, Cypress, Raytheon (chip), Integrated Telecom Technology (chip), Xylan, HP, Silicon Graphics, Advanced Telecommunications Modules, Trancell, SysKonnect, 3Com, NiceCom Ltd (bought by 3Com), ZATM, UB Networks, Swindon Silicon Systems (chips), SMC, LSI Logic (chips), Fibermux, Grand Junction, Tandem, Centillion, NetWiz, LANNET, CrossComm, AVIDIA, SNT, Hitachi, HyNEX, Cellware
FORE, SynOptics, Sun, TI (chip), NetEdge(FiberCom), Hughes, Cisco, 3Com, TransSwitch (chip), Alcatel, ATT, Fujitsu, General DataComm, GTE, Hughes, LightStream, NEC, NET, Network Systems, Newbridge, Northern Telecom, Telematics, TRW, DEC, Digital Link, Interphase, Network Peripherals, Odetics, Xyplex, PMC-Sierra (PM5346 chip), Olicom, Chipcom, Centillion Networks, RADCOM, AMCC (chip), Cypress (chip), SysKonnect, Allied Telesys, First Virtual, Bay, CrossComm, Hitachi
FORE, Cisco, Wellfleet, NetEdge(FiberCom), 3Com, TranSwitch (chip), Alcatel, Fujitsu, General DataComm, GTE, Hughes, LightStream, NEC, NET, Network Systems, Newbridge, Northern Telecom, StrataCom, SynOptics, Telematics, TRW, ADC Kentrox, Cascade, DEC, Digital Link, ODS, RADCOM, Brooktree (Chip), PMC (chip), Bay
TranSwitch (chip), Alcatel, Hughes, LightStream, Network Systems, Newbridge, Northern Telecom, StrataCom, Telematics, ADC Kentrox, Cisco, RADCOM, Brooktree (chip), PMC (chip), General DataComm, Bay, FORE
FORE, SynOptics, AMD (chip), Cisco, IBM, NET, General DataComm, Alcatel, General DataComm, GTE, LightStream, NEC, Newbridge, Cabletron, Digital Link, Interphase, NetEdge, Retix, Connectware, Chipcom, RADCOM, ZATM, Cisco, First Virtual, Bay
155MbpsUNI (Multimode)
155MbpsUNI (Type1/Type2)
155MbpsUNI (C5)
Micro Linear Corp (chip), SynOptics, Sun, Network Peripherals, Northern Telecom?, Connectware, Interphase, Trancell, SysKonnect, GEC Plessey (chip), Wolfson (chip), Bay, Cisco
Telco Systems, Cypress (chip)
Fujitsu, PMC-Sierra (chip), AMCC (chip), TI (chip), FORE, Hitachi, Maker, Cisco
ADC/Kentrox, Stratacom, NEC, Telecommunications Techniques, RADCOM, General Datacom
StrataCom, Telematics, FastComm, RADCOM, ADC/Kentrox, HyNet, Digital Link, NEC, General Datacom, Lightstream, Cascade, Telecommunications Techniques, FORE
FORE, Cellware
IBM (including chip), Chipcom, TranSwitch (ALI-25 chip), National Semiconductor, HP, ATM Limited, Cellware GmbH, Centillion Networks, Integrated Device Technology (chips), LSI Logic, Madge, Olicom, Silcom, Whitetree, Fujitsu (chip), First Virtual, Advanced Telecommunications Modules, Efficient Networks, ODS, Xircom, RADCOM, Apple, On Demand, Interphase, Adaptec, Rockwell, FORE, Bay, AVIDIA, SNT
Cornell University (prototype software), Connectware, IBM, Madge, 3Com
PMC-Sierra (chip), Cabletron, Cisco, BNR, Cascade, DSC
ATT, Newbridge, Northern Telecom?, Interphase, Silicon Design (chips)
Tut, UB Networks
Silicon Design (chips)
Firefly, Miro, Radius, TI (chipset), Sony (chip), Skipstone, Stewart Connector, Molex, Adaptec, Western Digital, IBM (chip), Symbios Logic, FujiFilm, Sun, Philips, Microsoft, Compaq, Intel, NEC (chip), Fujitsu
Apple, TI (chip), FujiFilm (chip), IBM, Adaptec, Molex, Skipstone, Digital VCR Alliance (consortium), Sony, Yamaha
TI (chip)
TI (chip)
Sony, Apple, Sun, Intel.
Apple, Compaq, HP, IBM, Unisys, Key Tronic, Mitel, Samsung, various PC companies


ATM "Asynchronous Transfer Mode"
a communications protocol that transmits data in 53-byte cells using switches and various line transmission technologies operating at different speeds. A lot of people associate ATM with high speeds, but really it is more of a framework for networking that is speed independent and the link that carries ATM data can be fast or slow.
DXI "Data Exchange Interface"
ATM Forum term.
"High Performance Parallel Interface", defined by ANSI X3T11.
HSSI "High Speed Serial Interface"
a 52Mbps interface between routers and DSUs, originally defined by Cisco and T#systems. It is also an ANSI standard.
MAN "Metropolitan Area Network".
Next "Near end crosstalk".
name for 802.3 method of negotiating between 10BASE-T and 100BASE-T.
OC-x "Optical Carrier level x"
A SONET term for an optically transmitted SONET signal at some particular speed. The base rate is 51.84Mbps. OC-1 runs at the base rate, OC-3 runs at 3 times the base rate, etc. Commonly planned rates are OC-1, OC-3 (155.52Mbps), OC-12 (622.08Mbps), and OC-48 (2.488Gbps).
PLCP "Physical Layer Convergence Protcool".
PMD "Physical Media Dependent". Term used to describe the layer of FDDI that determines the actual type of cable, etc. Also used in conjunction with other technologies, including ATM.
SMT "Station Management"
Network management protocol specific to FDDI.
SONET "Synchronous Optical Network"
A set of standard fiber-optic-based serial standards in North America. ATM runs as a layer on top of SONET (ATM also runs on top of other technologies). Developed by Bellcore and standardized by ANSI. Designed for telephone companies, for long-distance applications, but in the ATM world, being adapted to LAN uses.
SDH "Synchronous Digital Hierarchy"
Similar to SONET, but used outside North America. Some of the SDH and SONET standards are identical, in particular, the versions at 155Mbps and above interoperate. Standardized by the ITU-T. See SONET (above).
STM "Synchronous Transport Mode"
an SDH term.
STS-x "Synchronous Transport Signal level x"
a SONET term for an electrically transmitted SONET signal at some particular speed. Each STS level corresponds to an OC level (see OC-x above).
ATM Forum term for working group.
TAXI "Transparent Asynchronous Transmitter-Receiver Interface"
literally, the name of a chip from AMD originally designed to handle Multimode FDDI. Also, the popular name of two ATM interfaces developed by FORE by adapting the FDDI multi-mode physical layer and the chips AMD produced to support that part of FDDI. The slower of the two (100Mbps) was adopted by the ATM Forum and the faster of the two (140Mbps) was not.
See "ATM25".
UNI "User to Network Interface"
ATM Forum term.


(Note: a good resource is

American National Standards Institute
ANSI group developing standards for information processing.
old ANSI group within X3 that was developing standards for I/O interfaces.
ANSI X3T9.3 Committee
old name for ANSI X3T11 when it was part of X3T9.
ANSI X3T9.5 Committee
old name for ANSI X3T12 when it was part of X3T9.
ANSI X3T11 Committee
ANSI group standardizing HIPPI and Fibre Channel.
ANSI X3T12 Committee
ANSI group within X3T9 that standarized FDDI, PMD, SMF-PMD, and is standardizing TP-PMD and LCF-PMD.
International Electrotechnical Commission: not a usual standards bodies for LANs, but is involved in some technologies that are LAN-like.
Institute of Electrical & Electronic Engineers
IEEE 802 Group within IEEE that standardizes LAN technologies.
IEEE 802.3
Group within IEEE 802 that standardizes CSMA/CD LANs.
Former(?) group within IEEE 802 that studied the issue of a Gigabit version.
IEEE 802.3ab
Group within IEEE 802.3 to work on a Category 5 version of Gigabit Ethernet (1000BASE-TX).
IEEE 802.3b
Group within IEEE 802.3 to work on a Category 3 two-pair version of Fast Ethernet.
IEEE 802.3z
Group within IEEE 802.3 to work on a Gigabit version of Ethernet. Minutes, etc. at:
IEEE 802.6
Group within IEEE 802 that standardizes DQDB MANs.
IEEE 802.9
Group within IEEE 802 working on IS-LAN.
IEEE 802.12
Group within IEEE 802 working on 100VG-AnyLAN.
International Telecommunications Union (formerly called the CCITT)
ITU's Telecommunications Standards Sector.
ASTRAL "Alliance for Strategic Token Ring Advancement and Leadership"
Consortium of vendors working on new Token Ring technology developments. See
ATM Forum
Non-profit international industry consortium chartered to accelerate ATM acceptance & interoperability. Address: The ATM Forum; 303 Vintage Park Drive; Foster City, CA 94404-1138; 1(415)578-6860. Members include 3Com, 3DO, ADC Kentrox, AMP/ATM Systems, ATM Ltd, AT&T, SU-System, AWA Networks, Adaptec, Inc., AMD, AdvanceNet, Agile, Alantec, Alcatel, Allied Telesis, Ameritech, Apple, ascom Timeplex, BT Labs, Bear-Stearns & Co, Bell Atlantic, Bellcore, BellSouth, Bipolar Integrated Technology, Boeing, Bosch Telenorma, Broadband Technologies, Brooktree, Bull SA, CNT, COMSAT, CSELT, Cable & Wireless, Cablelabs, Cabletron, Centillion, Chipcom, Cisco, CompuServe, Cray Comm., Cray Research, CrossComm, Cypress, DSC Comm., Data Comm. Technology, David Sarnoff, DoD, Digiboard, DEC, Digital Link, Ericsson, E-Systems, Efficient Networks, Elec. & Telecom. Research, EXAR, FORE, France Telecom, Fujikura Technology America, Fujitsu, Furukawa Electric Tech., GPT Ltd., GTE Gov Systems, General DataComm, General Instrument, Graphics Comm., HP, Hitachi Telecom USA, Honeywell, Hughes LAN, IBM, IPC Info SYstems, Information Comm Inst Singapore, Integrated Device Technology, Integrated Telecom, Intel, Interphase, Joint Interoperability Test Center, KDD, Kalpana, LSI Logic, LANNET, Larscom, LLL, Lightstream, Loral Data Systems, MCI, MCNC, MFS, Madge, Microsoft, Mitel, Mitre, Mitsubishi, Motorola, Multimedia, NEC, NET/ADAPTIVE, NTT, NYNEX, National Inst. of Standards & Tech, National Semiconductor, NetEdge, Netrix, Network Communications, Network General, Network Peripherals, Netowrk Systems, Newbridge, Nokia, Northern Telecom, Novell, OKI, Olicom, OST, PMC Sierra, Pacific Bell, Philips, Proteon, QPSX, Quality Semiconductor, Racal-Datacom, Raynet, Raytheon, SGS-Thomson, SITA, Sandia National Labs, Scientific Atlanta, Siecor, Siemens, Sierra Research & Technology, Silicon Graphics, Silicon Systems, Sony, Southwestern Bell, Sprint, SMC, Stratacom, Sumitomo Electric, Summa Four, Sun, SuperNet, SynOptics, T3plus Neworking, TRW, TTC, Tekelec, Tektronix, Telco Systems, Telecom Finland, Telecom Italia, Telefonica I&D, Telematics International, Ltd., Telenex, Telia, Tellabs, Telogy, Telstra, Texas Instruments, RAD, Thomson-CSF, Toshiba, Transwitch, Trillium Digital Systems, TriQuint, US WEST, UB, Unisource Business, Unisis, VLSI Technology, VTT Information Tech, Valor Electronics, Verilink, Wellfleet Communications, Whitetree Network Technologies, WilTel, Xerox Parc, Xylan, Zeitnet, Zynrgy Group.
working group within the ATM Forum which deals with the physical technologies.
Desktop ATM25 Alliance
Coalition of 25 companies to develop the ATM25 specification and submit it to the ATM Forum. Members include IBM, Madge, Centillion Networks, Chipcom, Efficient Networks, First Virtual, Olicom, ODS, Xircom, Apple, On Demand, Interphase, LSI Logic, Fujitsu, Transwitch, Advanced Telecommunications Moldules Ltd., and Whitetree.
Fast Ethernet Alliance (FAE)
Now disbanded; was group of vendors working on the three variants of 100BASE-T/100BASE-F. Their stated reason for disbanding was that whereas the standards process was initially assisted by having a consortium to do the groundwork for the IEEE, with the ratification of the standard, the IEEE 802.3 committee has the process well in hand. Members included: 3Com, Cabletron, DAVID, DEC, Grand Junction, Intel, LANNET, National Semiconductor, SEEQ, SMC, Sun, Du Pont, Exar, IMC, JLP, LMC, Microlinear, NEC, Olympic Technology, Unisys, NetWorth, CNet, Cray, Hughes, Hyundai, Interphase, Montrose, Network General, Novell, Packet Engines Inc., ODS, Asante, Bay (both Synoptics and Wellfleet were members).
Full Duplex Switched Ethernet Consortium
Group of vendors that are working out the details of FDSE. Members include: Cabletron, Compaq, IBM, Kalpana, National Semiconductor, NCR, SEEQ, and Texas Instruments.
Fibre Channel Association
Group of vendors promoting Fibre Channel development, acceptance, and interoperability. See
Fibre Channel Systems Initiative (FCSI)
Group of vendors promoting a profile for using Fibre Channel for high-powered workstations. Members include: HP, IBM, Sun.
HIPPI Networking Forum
Consortium of vendors (initially 11) to promote the use of HIPPI. Members include: AMMC, Avaika, Broadband, E-Systems, Essential, IBM, Loral Defense, Los Alamos National Lab, Maximum Strategy, Methode, Myriad Logic, NSC, NetStar, PsiTech, Silicon Graphics, Triplex Systems, TRW, University of Illinois, and University of Minnesota.
100VG-AnyLAN Forum
Group of vendors trying to accelerate 100VG-AnyLAN acceptance & interoperability. Charter members include ATT, HP, IBM, ODS, Proteon, UB, Wellfleet. Address: North Highland,s CA, Phone number: 1(916)348-0212.
University of New Hampshire InterOperability Lab
Organization designed to improve the operation of hetrogeneous networks. See
IncAlliance "Isochronous network communciation alliance"
Group of vendors working together to foster and manage the deployment of isoEthernet. Members include Apple, Ascom-Nexion, AT&T, DataBeam, Dialogic, Ericsson, Future Labs, IBM, ITT, Incite/Intecom, Luxcom, MCI, National Semiconductor, Pacific Bell, Primary Rate, Quicknet, Siemens/Rolm, VCON, Zydacron.
Digital VCR Alliance
Group of vendors defining standards for digital VCRs and camcorders. They are including FireWire interfaces in the standard.
Gigabit Ethernet Alliance (GEA)
Group of vendors working together to help drive the standarization of Ethernet at 1Gbps data rates. Its 100-plus members include 3Com, Acacia, Ancor, Apple, Bay, Cabletron, Cisco, Compaq, Cypress Semiconductor, Digital, D-Link, FORE, Granite, HP, IBM, Intel, Kingston Technology, LANart, Level One Communications, LSI Logic, Lucent, Madge, MediaWise, NBase, National Semiconductor, Ornet, Packet Engines, Shiva, Silicon Graphics, Sun, TI, UB, VLSI Technology XaQti, XLNT Designs, ZNYX, Netcom Systems, Network General. See
1394 Trade Association
Group of vendors promoting the use of FireWire a.k.a. IEEE 1394. Membership includes more than 45 companies, including Apple, Microsoft, Sun, Compaq, and Intel. See a.k.a.
CIF Alliance
Group of vendors promoting CIF. Members include 3Com, 3M, Agile, Apple, Bay, Bell Atlantic, Brooktree, Cisco, Connectware, Cornell University, First Virtual,IBM, Madge, Matsushita, NSF, PMC-Sierra, Stratacom, Sun laboratories, UC Davis, and Whitetree. See

Selected Major Vendors' Activities

(note: virtually all deal with 10BASE-T and older Ethernet styles and multimode PMD FDDI; Many are members of the ATM Forum, but I don't have a list of members)

Computer Vendors

primary proponent of 4Mb Token, 16Mb Token, SwToken, FDToken, ATM25, 100Mb Tkn, 1Gb Token; active in TP-PMD, 100VG-AnyL, SwEthern, FDSE, IS-LAN, 100MbpsUNI, 10BASE-FB, FibreChan, 10BASE-FB, SDDI, HIPPI-PH32, 100BASE-TX; consortiums: Full Duplex Switched Etherent, Fibre Channel Systems Initiative, 100VG-AnyLAN Forum, ATM Forum, HIPPI Networking Forum, Gigabit Ethernet Alliance.
primary proponent of SwFDDI and FFDT; active in SMF-PMD, TP-PMD, 100BASE-TX, 100BASE-T4, 100VG-Anyl, FDSE, STS3cUNI, DS3UNI, HIPPI-PH32; consortiums: Desktop ATM25 Alliance, ATM Forum.
active in 100BASE-TX, FDSE, FibreChan, STS3cUNI, C5 155MbpsUNI, HIPPI-PH32, IS-LAN; consortiums: Fibre Channel Systems Initiative, ATM Forum, Gigabit Ethernet Alliance.
primary proponent of LocalTalk; active in IS-LAN, ATM, ATM25, 100BASE-T; consortiums: ATM Forum, Desktop ATM Alliance.
primary proponent of 100VG-AnyL; active in SwEthern, FDSE, FibreChan, ATM25, HIPPI-PH32; consortiums: Fiber Channel Systems Initiative; 100VG-AnyLAN Forum, ATM Forum, Gigabit Ethernet Alliance.
active in FDSE, 100VG-AnyL, 100BASE-T; consortiums: Full Duplex Switched Ethernet Consortium, Gigabit Ethernet Alliance.

Software vendors

active in IS-LAN; consortiums: ATM Forum.
consortiums: ATM Forum.

Datacomm equipment vendors

primary proponent of PACE; active in TP-PMD, 100BASE-TX, 100BASE-T4, SwEthern, FDSE, STS3cUNI, DS3UNI, 100VG-AnyL, ATM, 1000BASE-X; consortiums: ATM Forum, Gigabit Ethernet Alliance.
active in TP-PMD, 100BASE-TX, 100BASE-FX, 100BASE-T4, SwEthern, FDSE, FDFastE, STS3cUNI, 100MbpsUNI, 155MbpsUNI; consortiums: ATM Forum.
active in FDFDDI, TP-PMD, 100BASE-TX, SwEthern, FDSE, 100MbpsUNI, 1000BASE-X; consortiums: Full Duplex Switched Ethernet Consortium, ATM Forum, Gigabit Ethernet Alliance.
primary proponent of TP-PMD; active in 100BASE-T, STS3cUNI, DS3UNI, E3UNI, 100MbpsUNI, 100VG-AnyL, 1000BASE-X; consortiums: ATM Forum.
active in 100BASE-TX, DS3UNI; consortiums: 100VG-AnyLAN Forum, ATM Forum, Gigabit Ethernet Alliance.
(see Wellfleet & Synoptics; I'll add items here when I see explicit news releases about Bay's plans) active in 1000BASE-X. Consortiums: Gigabit Ethernet Alliance.

Chip manufacturers

active in 100BASE-TX, 100BASE-T4, FDFastE; consortiums: ATM Forum.
consortiums: ATM Forum.
Texas Instruments
active in 100BASE-T, 100VG-AnyL, STS3cUNI, FDSE; consortiums: Full Duplex Switched Ethernet Consortium, ATM Forum, Gigabit Ethernet Alliance.

Some History

The idea spurred three proposals: HP's 100BASE-VG which doesn't use CSMA/CD, and two CSMA/CD proposals, 4T+ and 100BASE-X. Some IEEE 802.3 members objected to 802.3 working on a non-CSMA/CD proposal since "by definition", that group worked on standardizing CSMA/CD networks. That held up standardization efforts for a while, but finally a new group 802.12 was formed. All three proposals still live: 100BASE-VG was expanded to also carry Token-Ring style packets and renamed 100VG-AnyLAN; 100BASE-X is now called 100BASE-TX, and 4T+ is now called 100BASE-T4.
Several vendors introduced priorietary methods of running FDDI over coax & twisted-pair. Crescendo's CDDI was adopted by ANSI with modifications (of course) as TP-PMD and for a while Crescendo still used the term CDDI for their standard products, but Cisco (who bought Crescendo) has dropped the trademark on the term so the industry can apply it to TP-PMD. Other methods were DEC's methods for running it over STP and Ethernet ThinNet-style coax, IBM's SDDI for shielded twisted-pair, and the other UTP competitor: TP-FDDI. SDDI is still marketed.
Proposals date back at least to mid-80s. The technology grew out of efforts for a new type of switching for both voice and data for nation-wide networks, and the famous 53-byte cell was originally proposed to be even smaller by those interested in voice transmission. Several vendors proposed it as a future LAN technology and then helped form the ATM Forum to push the technology. The ATM Forum seems to be a model for a new type of organization: not a standards organization, but a group of vendors who write "protocol definition documents" and propose them to the standards bodies, hoping to get them through without any signficant changes, and remaining comfortable that they can sell it well enough to create a defacto standard if need be. An interesting result is that whereas standards bodies are very open in their standards-defining process, these organizations are private and do their writing & discussing in secret. The organizations are typically funded by membership fees, and few companies other than network equipment vendors are willing to put up the money.
ATM25 vs 51MbpsUNI vs STS-1/2
These are competing standards for the low end desktop ATM, i.e. UTP. IBM pushed ATM25 and some other companies pushed a 51Mbps standard. The Forum initially decided picked the 51Mbps standard, rejecting the other so as to maintain focus, but in February 1995, after the ATM25 porposal was resubmitted as well as another competitive proposal (PCM-Sierra's STS-1/2), the Forum chose to move forward with ATM25.
Gigabit Ethernet
Efforts to develop and standardize around mid-90s, borrowing technology from FibreChannel. The high speed and Ethernet's small packet size and the use of CSMA/CD require the distances supported be very small, and switching is becoming more common, so there has been some controversy in whether CSMA/CD should be supported, or whether the packets can be extended or batched in such a way to support CSMA/CD at a useful distance.


Issues: Factors that will decide the winners

  1. Customer demand for more speed at a lower price. Full-motion video is on the way & file servers remain popular.
  2. How quickly various products are brought to market.
  3. How quickly various standards stabilize.
  4. Customers' installed wiring plants. Something that can run on any line currently running 10BASE-T will have some advantage.
  5. Customers' installed NICs. A couple of the technologies require no change.
  6. Cost of any new equipment or media that is needed.
  7. Interoperability between existing customer equipment: If customers need to connect an X server to a Y client, and perhaps necessarily through a Z piece of networking equipment, they will need a technology supported by X and Y (and perhaps Z).
  8. Packet formats: weighing the advantages over simplified bridging due to using a technology that uses a current packet technology.
  9. Need for new network software, possibly up to application level. ATM might become "just another LAN technology" or might revolutionize everything and take over, eliminating traditional routers. The latter would require considerable change in a lot of software. Who's writing such software? Any real efforts to propogate it to as many end nodes as currently use 10BASE-T?
  10. Distance limitations: high-speed CSMA/CD is OK for short distances, but CSMA/CD would have to be "disabled" to run higher speeds over some distances. In particular, 100mbps CSMA/CD with Ethernet-sized packets (512 bits minimum) has a radius (maximum distance) of 250 meters including two repeaters. Without repeaters, it can be extended to about 400 meters.
  11. Ability to offer expensive technology only to needed locations efficiently: customer sites often have a relatively small percentage of users who need higher-speed technology, and they aren't always co-located.
  12. Potential to integrate LANs and WANs. ATM has been projected to blur the distinction, making things easier.
  13. Danger of integrating your LANs too closely with those of other organizations: do you want, in effect, your competition sharing your LAN? All the high-schools and colleges throughout the world? Lots of sites set up extra firewalling before attaching their LAN to the present Internet.
  14. Customers' and vendors' desire to pick a long-term winner: customers listen to vendor announcements and vendors ask customers what they are likely to buy, both wait a while, and worry about what to commit to. Predicting the future is never easy, and never certain.

The hypothetical perfect technology

  1. No translation bridging required
  2. Uses current NICs
  3. Supports large packets
  4. No doubt of success & long future
  5. Huge market & lots of vendors
  6. Low cost
  7. Runs long distances
  8. Runs over 2PC3
  9. Runs over WANs as easy as LANs
  10. No high-cost/high-maintenance routers acting as bottlenecks