WO1990016026A1

WO1990016026A1 - Local area network bridge module

Info

Publication number: WO1990016026A1
Application number: PCT/US1990/003454
Authority: WO
Inventors: Robert A. Craven
Original assignee: Halley Systems, Inc.
Priority date: 1989-06-20
Filing date: 1990-06-19
Publication date: 1990-12-27
Also published as: AU5951690A

Abstract

A bridge module (1) for token ring local area networks (25, 30) is disclosed. The bridge module utilizes content addressable memory (10, 12) and negative filtering to provide extremely rapid communication between two or more token ring local area networks (10, 12) using source routing techniques. A portion (76) of the content addressable memory is dynamically allocated to assist in providing routing information for non-source routed messages, thereby enabling the bridge module (1) to forward both source and non-source routed messages in a manner transparent to the receving computer.

Description

LOCAL AREA NETWORK BRIDGE MODULE

FIELD OF THE INVENTION This invention relates to the field of computer interconnection. More specifically, it relates to a bridge module for communication between two or more local area networks.

BACKGROUND OF THE INVENTION Networks of computers are known. The earliest example of computer networks is several users sharing a single computer. This arrangement, known as time- sharing, maximized use of early expensive main frame computers. Recently, the explosive proliferation of relatively inexpensive personal computers has created new motivations for linking computers. These include the sharing of large data bases, information exchange between individual personal computers, and collaboration between various users on a single project.

One common computer network is the Local Area Network or LAN. A LAN comprises two or more computers generally located near to one another and linked together. Such LANs may connect all the computers in a single office or building; typically these LANs serve the needs of a single organization. They allow members of the organization to send messages to one another, access the same programs and data bases, and share storage facilities and printers. Such networks are often based on the use of shared coaxial cable or twisted pairs of wires.

Two topologies have been used to create LANs. One is a tree, of which the Xerox Corporation's Ethernet is an example, and the other is the ring, of which IBM Corporation's token ring is an example. Both topologies and their methods of operation are known. As knowledge of token ring LANs is necessary for an understanding of this invention, a brief description of such LANs follows. In a token ring IAN, access to the LAN is by means of a token (a short sequence of bits) that circulates continuously around the ring. In order to transmit data, a machine interface must first wait for the token to pass. It removes the token temporarily and inserts an addressed packet (of up to a specified maximum length) onto the ring. The packet is received and copied by the intended recipient before the message makes a full circle around the ring. When the message makes a complete circle, the sending station removes the message and reinserts the token onto the ring, enabling the next sender to "take" the token and repeat the procedure. An important aspect to note is that each message passes through each computer in the complete ring. Thus, each computer functions as a "repeater" for the message.

All LANs have certain physical limitations regarding the maximum wire length of the LAN and the maximum number of users which may be connected to a single LAN. These limitations are the result of propagation delays over long networks, noise, competition among the various computers for use of the LAN, etc. Although this maximum size varies widely among the different types of LANs, many organizations have found that their LANs have reached the maximum permissible size.

Often, even when not mandated by physical limitations, it is desireable to have two or more smaller LAN's than one large LAN. For example, consider a LAN connecting the computers in the accounting department of --£>- a corporation. To preserve data security, workers in other departments should not be allowed to access the accounting department's IAN. Placing the accounting department's computers on a separate LAN solves this problem. Similarly, division into a plurality of LANs can facilitate user accountability and network manageability.

A consequence of the physical size limitations and the other stated considerations is the need for a bridge between LANs. Such bridges result in an effective increase in the size of the IAN and are known. To control the traffic between token ring LANs with such bridges, two main strategies of inter-IAN traffic routing have evolved. The first strategy, called "Source Routing", is used by IBM. In this approach, the path that an internetwork message must take to reach its destination is embedded in the message itself. The route is discovered by broadcasting the message to all rings, the message being addressed to an intended destination address. As the message passes through the bridges, each bridge adds itself to the routing information within the message. The destination computer may receive as many of these broadcast messages as there are alternative routes between the computers. The destination computer simply responds to all of the messages that it receives and does so by using the routing information that has accumulated during the path out. The receiving computer tells the bridges to use the routing field in reverse order by changing the direction bit in the routing field. This allows the bridges to pass the message back to the sending computer in exactly the reverse path by which it was transmitted to the destination computer. The original transmitting^" computer then accepts the first response that it receives, stores the routing information from this message for subsequent use and discards any other paths. The assumption is that the fastest path is the best path. The second internetwork routing strategy is called "Non-Source Routing". After a determination is made that the destination computer is not on the local IAN, the bridge uses the message's destination address to perform a table lookup. This table lookup returns with the routing information necessary to send a message to the destination computer. This approach has the advantage of being protocol independent, but makes it necessary for a bridge to use two fundamentally different strategies to route different kinds of messages, even though both may occur on the same subnet.

SUMMARY OF THE INVENTION The present invention provides a bridge module for token-ring LAN's which can forward messages using either "Source Routing" or "Non-Source Routing" strategies, thereby controlling internetwork messages.

In one embodiment the bridge module connecting the LANs comprises a protocol handler, a buffer memory, two identical IAN interface modules with two associated identical content-addressable memory (CAM) units.

When a message is sent from a first computer on a first token ring LAN to a second computer on a second token ring IAN, the bridge module, which interconnects the first and second IAN, performs several tasks almost simultaneously.

As the message begins to pass through the bridge module, a copy of it is begun, the copying process being a continuous one as the message passes through the bridge and the copy as it is made being stored in the bridge module's buffer memory. The original message, as it is being copied, is also being retransmitted onto the first IAN. As the message proceeds through the bridge, as soon as the Protocol Handler has detected the second through fourth word of the message, which contains the message's Destination Address ("DA"), the DA is sent to the CAM unit dedicated to the first IAN.

The message's DA, which comprises 48 bits, is compared with the contents of the first CAM unit. The content of the CAM is a list of the DAs of all computers in the first IAN. If the DA of the message is found in the CAM, the message is for a computer on the first IAN and the bridge module clears its buffer memory of the partial message contained therein and continues retransmission of the remainder of the original message. On the other hand, if the DA of the message does not appear in the first CAM, the message is for a computer in the second or subsequent IAN. In this case, the module continues to copy the message into its buffer. After the copy is complete, it is sent to the bridging module's

Protocol Handler for retransmission onto the second IAN. The comparison of the DA with the contents of the CAM is an extremely fast operation, requiring roughly 100 nanoseconds. These fast comparisons allow the bridge module to utilize the entire DA for routing messages, in real time. Additionally, as the decision to forward the message is based on "negative filtering" — only the absence of the DA from the first CAM causes the message to be transmitted to the Protocol Handler, not its presence —, the limited amount of storage in the CAM, only 256 48-bit words, creates few problems, as the number of LANs containing 256 nodes is small.

An additional feature of the present invention is its ability to work with either source or non-source routed messages. When a message is not source-routed, prior art bridges needed to use all 48 bits of the DA to access a table of active computers to find the routing necessary to reach the particular computer. The time needed to complete a 48-bit table lookup is very long in computer terms. In the present invention, the Protocol Handler makes a ones-complemented copy of the non-source routed message's DA. This complemented copy of the DA, which cannot be a valid address for a computer on the first IAN, is compared to the contents of the CAMs. The CAM, in addition to the DAs of active computers on the first IAN, also can contain a list of ones-complemented addresses for table lookups. More precisely, the CAM has a ones-complemented list of actively used computers not on the first IAN. If the destination is a valid one (not on the first IAN) , the position in the CAM of the ones - complemented address of the particular destination computer will be returned from the comparison operation. The position has a value which ranges from 0-255 and is used to access a table containing routing information to obtain the routing information for the particular computer having that DA. As only 100 nanoseconds are needed to obtain the offset, the use of this bridge module results in a large speed improvement in routing non-source routed messages.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram of the preferred embodiment of the present invention; Fig. 2 shows the various protocol layers used in IAN networks;

Fig. 3 shows the format used for a source routed message; -T -

Fig. 4 is a detail of the IAN interface shown in Fig. 1;

Fig. 5 is a flow chart showing the process for determining whether a message is an inter- or intra- IAN message;

Fig. 6 is a flow chart showing how source and non-source routed messages are forwarded to other LANs;

Fig. 7 shows two token ring LANs linked by one bridge module; and Fig. 8 shows three token ring LANs linked by two bridge modules.

DETAILED DESCRIPTION Physical Configuration A block diagram of a preferred embodiment of the present invention is shown in Fig. 1. LAN bridge 1 comprises two identical content addressable memory ('CAM') units 10 and 12. The CAM units are each comprised of a AM99C10 CAM chip with supporting circuitry. The CAM can store 256 48-bit words and a 48- bit input word can be compared against all 256 words stored in the CAM in a single 100 nanosecond cycle. CAM units 10 and 12 are respectively coupled to IAN interface units 14 and 16. The IAN interface units are comprised of TMS380 LAN adapter chipsets made by Texas Instruments. A detail of these interface units is shown in Fig. 4 and the units' operation is discussed below. IAN interface units 14 and 16 are in turn coupled to multi-access units 18 and 20, respectively. The construction and operation of these units is known and they link the IAN interface units to respective LANs 25 and 30.

A central processing unit 40 with 640K RAM main memory 42 is coupled to the CAM units and IAN interface units by means of bus 44. In one embodiment an IBM PC-AT compatible computer is used as the central processing unit 40.

In the preferred embodiment both IAN interface units and CAM units are mounted on the same circuit board. This arrangement is not necessary for the proper operation of this invention and many other arrangements can be readily visualized.

Further details of the construction and operation of IAN interface units 14 and 16 can be found in the "TMS 380 Adapter Chipset User's Guide" and the

"TMS 380 Adapter Chipset User's Guide Supplement", both published by Texas Instruments, and copyrighted by IBM and T.I. These manuals are incorporated in this description in their entirety. As the present invention is designed to provide bridging between token ring LANs using either source or non-source routing, several computer industry protocols and standards are of necessity incorporated in this invention. The Open Systems Interconnection (OSI) model is a conceptual network structure defined by the International Organization for Standardization. As shown in Fig. 2, networks are partitioned into a series of layers, numbered 1 through 7, to reduce their design complexity. These layers are hierarchical with each layer being built upon its predecessor, thus shielding each layer from the details of how the services from the other layers are implemented. For units to communicate at any given layer, the operating protocols of the components must correspond from the first layer up to and including the layer where the units will operate to insure full interoperability. The layers of concern in the present invention are the physical layer 60 and data - 5 - link 62. Although current networks do not conform fully with the OSI model, they are analogous in most respects.

Physical layer 60 defines the mechanical and electrical connections between the various units on the IAN. Data Link layer 62 defines the way data is formatted for transmission and how access to the network is controlled. This layer has been separated by the IEEE 802 Standards Committee into two sublayers: the Medium Access Control (MAC) sublayer 623 and the Logical Link Control (LLC) 624 sublayer. The present invention operates up to the MAC sublayer, which is lower than the LLC. IEEE standards 802.2 govern the MAC sublayer and 802.5 governs the physical layer.

In Fig. 3, a frame (message) which is used to send data on a IAN is illustrated. Although each area of the frame performs an important function, this invention is primarily concerned with the source 70 and destination addresses 72 (SA and DA, respectively) . Source routing, which was described in the summary of this invention, involves an IBM extension to the IEEE 802.2 standards and is indicated by turning the first bit of the SA to 1. If this occurs, the bridge knows that routing information 76 needed for the present message is stored in an added field placed between the SA and the information field. In non-source routing, which was also described in the summary, the DA is used to access a table containing the routing information.

OPERATION OF THE BRIDGE Initialization

A standard feature of token ring LANs is that one of the computers on the IAN becomes the active monitor using a known process called monitor contention. Every seven seconds the active monitor polls the ring to - \ o determine which other computers are on the ring. The active monitor transmits an Active Monitor Present (AMP) message. As each computer receives this message they in turn transmit a Standby Monitor Present (SMP) message. As each message proceeds around the ring and goes through the bridge module, the SA from each message is stored by the bridge's control processor in the CAM unit related to that IAN. Referring to Fig. 6, step 202 indicates the bridge module receiving a frame over the LAN. Step 204 shows the bridge testing the frame to see if it is either an AMP or SMP. If it is, it is added to the bridge's CAM at step 206. Thus, every seven seconds the CAM is updated with the SA from all active computers on the IAN. This process occurs on both rings 25 and 30 in Fig. 1, so CAM units 10 and 12 are both updated every seven seconds.

Another important initialization step involves the Duplicate Address Test (DAT) . When a computer is coupled to a LAN, it transmits a self-addressed message. In other words, the message's source and destination addresses are identical. If the address assigned to the new computer is unique, the message will proceed around the IAN without being copied. If it is copied, one of the bits in frame status byte 74 of the message (See Fig. 3) will be set to one. The original sender will, when the message returns to it, detect the changed bit and know that its assigned address is not unique. This test is repeated three times when a computer tries to access the.IAN. If it is copied any time other than during the first attempt, the computer is not allowed on the ring. Although this test is necessary to prevent access to the IAN by computers having addresses which are the same as the addresses of other computers on the IAN, when a bridge module is coupled to the IAN, a special problem manifests itself. As the new computer attempts - u to access the IAN, it transmits its DAT message. The bridge does not have the address of this computer in its CAM as the new computer was not "polled" earlier. Consequently the bridge copies the message, forwards it to the other LANs, and switches the appropriate frame status bit to indicate that the message was copied. The computer trying to access the IAN receives its DAT message back, with its frame status byte on. Consequently, the computer knows its message was copied and it finds itself unable to access the IAN. This set of events could continue indefinitely. Fortunately, in token ring IAN architecture the first word of the DAT is unique. This uniqueness allows for a check routine in the present invention which overcomes the problem just described . As the solution is part of the operation of the bridge module, it is described below.

Referring now to Figs. 4 and 5, as a message enters the bridge module it is buffered, a few bits at a time, into the Protocol Handler 64 (Fig 4) . After a 16- bit word is assembled in the Protocol Handler's buffer, it is transmitted over IAN Adapter Bus 65 to the buffer RAM 66 in Communications Processor 67. The bridge module monitors the interface between the Protocol Handler and the Communications Processor. As the first word is received and copied at step 102 in Fig. 5, the bridge module checks to see if the word matches the unique first word which prefaces all DAT messages. This is shown as step 104 in Fig. 5. If it does, the XFAIL signal is asserted which in turn causes the bridge module to ignore the rest of the message and clear its copy of the message (step 106, Fig. 5) . This process allows new computers to access the net, without their DAT messages being copied and thereby being denied access to the LAN. If a message is not a DAT message the bridge unit continues to monitor and copy the traffic between Protocol Handler 64 and buffer RAM 66. As each word of the message's DA is received, it is copied and transmitted to the first CAM unit. This process is shown as steps 108, 110 and 112 in Fig. 5. After all 3 words of the DA are sent, the comparison described below takes place.

In the CAM unit the DA of the current message is compared to the list of active computers on the LAN stored in the CAM. This occurs at step 114 in Fig. 5. If a match occurs, this means that the DA is that of a computer on the local IAN and the logic associated with the CAM unit asserts the XFAIL pin on the TMS38021 protocol handler. In this case, the presence of the DA in the CAM indicates that a copy of the rest of the message is not needed. Although the TMS38021 protocol handler may continue to copy the message based upon its own criteria, the CAM unit will not cause a copy to be made and merely waits for the next message from the LAN. This process occurs at step 118. The use of XFAIL based on detecting the presence of the DA is part of the realization of the negative filtering concept.

If the DA is not in the CAM for the first LAN, the XMATCH pin on the TMS 38021 is asserted and a copy of the frame will be made at step 116. The Protocol Handler then orders a comparison of the DA with the contents of the second CAM chip in the bridge module. See step 208 in Fig. 6. If the DA is in the second CAM, the copy of the message present in the Protocol Handler's memory is queued for transmission onto the second IAN at step 210. In this special case, there is no necessity of knowing whether the message is source-routed or non-source routed as the location of the destination computer is known to the bridge module.

Referring to Fig. 7, if computer 304 transmits a message to computer 306, the comparison operation indicated by step 114 in Fig. 5 will find that the DA of computer 306 is in the CAM unit dedicated to IAN 310 and the bridge module 300 will stop copying the frame and merely pass it on to computer 306. If computer 304 transmits a message for computer 326 in IAN 320, the comparison operation will initially return a negative result. Upon performing the test indicated by step 208 in Fig. 6, the presence of computer 326's DA in the CAM unit dedicated to IAN 320 will cause the message to be transmitted onto IAN 320. This forwarding occurs regardless of whether the frame is source or non-source addressed.

A special procedure for messages intended for the bridge module itself is shown in Fig. 6 at steps 212 and 214. At step 212 a test is performed to see if the frame is intended for the bridge module. If it is, the frame is forwarded to the bridge module at step 214.

The bridge module checks to see if the frame is source routed at step 216. If it is, routing field 76 of the message (see Fig. 3) is examined at step 218. If the segment number (a 2-byte number in hexadecimal notation) of the bridge module performing the comparison is not found in the frame, the frame is discarded at step 222. If the segment number of the bridge module is found, the bridge forwards the message according to source routing protocol at step 220.

In the case where the computers use source- routing, the absence of the message's DA from either of the bridge module's CAMs indicates that the message must be forwarded according to source-routing protocol. This - \ +- procedure was described in the summary and simply means that the bridge module forward the message according to known IBM protocols. Whether or not this requires any the bridge module to perform any additional actions depends upon the evaluation of routing field 76's information.

The proper forwarding of non-source routed messages is more complicated. In a source routed frame, the first bit of the source address is set to 1. In a non-source routed this first bit is set to 0. At step

224 the Protocol Handler checks to see if the frame is a non-source routed frame. If the frame is not a non- source routed message, it is discarded at step 226. At step 228, the ones-complement of the DA is taken and compared to the contents of the first and second CAMs. Each CAM has, in addition to its list of DAs for computers on its IAN, a list of computers for whom routing information is stored in the 64OK RAM. The CAM unit list is a list of the ones-complemented DAs of those computers. The ones-complemented addresses are in a form which cannot be mistaken for a valid DA. During normal operation, if the ones-complemented address is found in the CAM at step 230, the comparison operation detects it and returns a number which indicates its position (0 through 255) in the CAM at step 232. This number is used as an offset which allows the Protocol Handler to access the RAM table containing the routing information for the computer with this DA. The routing information obtained is then used to forward the message at step 234. Initially, if the current message is the first message for this destination computer, neither CAM will contain the ones-complemented DA. In this event an entry of this ones-complemented address is placed in the first CAM and a table entry, which will be blank initially, is created in the CAM at step 236. An All Rings Broadcast message is transmitted according to IBM protocol to find the bridge module adjacent to this DA at step 238. The first positive response, which now contains routing information to the indicated bridge module, is used as the primary route and entered into the routing table, as is indicated by step 240. A second positive response becomes the alternative route, used only when the primary route becomes unavailable. Subsequent responses are ignored.

As this process is somewhat time consuming, the initial message is now discarded. When it is retransmitted by the source computer, an envelope is placed around the message, the envelope being source- routed to the just÷discovered nearest adjacent bridge module. The envelope conforms to the standard source- routed format. In the sample frame of Fig. 3, the address of the adjacent bridge module is used for DA 72, SA 70 is the sending bridge module's address, and the entire original message becomes the information field.

All other parts of the frame are standard and require no special treatment. The message now becomes a bridge to bridge communication. When the destination bridge receives the message, it strips off the envelope and forwards the original message to the intended destination computer.

Fig. 8 shows three separate LAN's linked by bridge modules 400 and 450. If a source-routed message is transmitted from computer 432 to 426, bridge module 450 first compares the DA of the message to the DAs in the CAM associated with IAN 430. When the DA is not found in that CAM, the CAM associated with IAN 420 is searched. The DA of computer 426 will be found therein and the message transmitted onto IAN 420. If computer - \b- 432 attempts to send a message to computer 412 on IAN 410, similar comparisons will fail to find the DA in bridge module 450. Consequently, bridge module 450 will transmit the message onto IAN 420, where it reaches bridge module 400. Bridge module 400 will receive the message and perform the same two comparisons that bridge module 450 performed. The second comparison will find the DA of computer 412 and the message will be placed on IAN 410 by bridge module 400 from whence computer 412 will eventually receive it.

If the message from computer 432 to computer 412 had been non-source routed, the initial comparisons performed by bridge module 450 would be identical. However, as the message would be for a computer on a non- adjacent IAN, the ones-complement of the DA would be used in the previously described fashion to obtain the proper route information to reach computer 412. An envelope would be placed around the message with the DA of the envelope being that of bridge module 400. This procedure results in complete transparency of forwarding operations to the user regardless of whether the message is source or non-source routed.

The preceding specification describes with detail and specificity a bridge module for linking LANs which uses CAMs in a "negative filtering" format and which operates transparently regardless of whether the message is source or non-source routed. Although a particular embodiment has been described, several variations of this embodiment are envisioned. For example, the fast filtering provided by the CAM can be used for security. The CAM could be preloaded with addresses to prevent off-net access. Thus special nodes and servers would be well protected. The fast filtering could also be used for access control on file servers, or - \ι - communication servers. Additionally the bridge module could be used to link token ring LANs via a wide area network, token ring LANs to token ring LANs via an Ethernet IAN, Ethernet LANs to one another via wide area networks and Ethernet LANs together via a token ring IAN. The substitution of networks constructed using ISDN and FDDI protocols and hardware is also envisioned. These various modifications and changes may be made to the invention without departing from the broader spirit and scope of the invention as set forth in the appended claims. The specification and drawings are, accordingly, to be regarded in an illustrative rather than in a restrictive sense.

Claims

- \ e> -Claims;

1. A bridge module for linking at least a first and second local area network comprising: central processing means; a first and second local area network interface means each coupled respectively to the central processing means and the first and second local area networks; and first and second content addressable memory means coupled to the first and second local area network interfaces.

2. The bridge module of Claim 1 wherein a main memory means is coupled to the central processing means.

3. The bridge module of Claim 2 wherein a first and second multi-access means is coupled to the first and second local area network interfaces and to the first and second local area network.

4. A bridge module for linking at least a first and second computer network comprising first and second content addressable memory means for storing respectively the addresses of the computers on the first and second computer networks; first and second network interface means coupled respectively to the first and second content addressable memory means and the first and second computer networks for monitoring messages transmitted on the first and second computer networks, copying the messages and providing the content addressable memory means - \ °\ - with the destination addresses of the messages transmitted on the first and second computer networks; and central processing means coupled to the first and second network interface means for controlling the interaction between the interaction between the first and second computer networks, the first and second content addressable memory means, and the first and second network interface means.

5. The bridge module of Claim 4 wherein the central processing means has a main memory means coupled thereto for storing the programs which operate the central processing means.

6. The bridge module of Claim 4 further comprising a first and second multi-access means coupled to the first and second computer network and the first and second network interface means for providing an electrical and mechanical interface between the network interface means and the computer networks.

7. The bridge module of Claim 5 further comprising a first and second multi-access means coupled to the first and second computer network and the first and second computer network and the first and second network interface means for providing an electrical and mechanical interface between the network interface means and the computer networks.

8. A method for linking at least a first and second local area network, the local area networks -re^¬ transmitting frames having at least a destination address, the method comprising the steps of monitoring the first and second local area networks for the transmission of frames; reading each frame as it is transmitted over its respective local area network; comparing the destination address of each frame with the contents of a content addressable memory assigned to the network which stores the destination addresses used by that network; and forwarding the frame to the network which did not originally transmit the frame when the destination address of the frame is not in the content addressable memory assigned to that network.

9. The method of Claim 8 wherein the step of reading each frame further comprises checking the frame to see if the frame is in a predetermined form indicating that it is a test frame; and retransmitting the frame onto the network which originally transmitted it if the frame is a test frame.

10. A method for determining internetwork transmission of frames, the frames being comprised of at least a destination address for the frame and data, comprising storing the destination address of all frames not requiring internetwork transmission; comparing the destination address of a new frame to the stored destination address; and - l - transmitting the new frame internetwork when the destination address of the new frame is not the same as the stored destination addresses.

11. A bridge module for transmitting information between at least a first an second local area network of at least two computers, the information taking the form of at least a source address, a destination address and data; comprising: first and second means for storing the destination address of the computers coupled to the at least first and second local area networks, the memory means being addressed by the destination address; first and second local area network monitor means coupled respectively to the first and second memory means and the first and second local area networks for copying the information transmitted on the respective local area networks and providing the destination address of the information to the memory means; central processing means coupled to the memory means and monitor means for determining when the destination address of the information is not in the memory means and for controlling the forwarding of information when the information's destination address is not in the memory means.

12. A method for transmitting data frames between a plurality of computer networks, each network having only one common point between it and any other network and the data frames comprising at least a -Z2-- destination address and information, the method comprising the steps of storing at the common point between two networks the destination address of the computers which comprises the two networks; determining if a data frame transmitted by either of the two networks has a destination address differing from the destination addresses stored at the common point; and forwarding the data frame to the network which did not originate the data frame if the destination address of the data frame is different from those stored at the common point.

12. The method of Claim 12 wherein the common point comprises a bridge module.

13. The method of Claim 13 wherein the destination addresses are stored in a content addressable memory.

14. The method of Claim 12 wherein the information in the data frames may or may not contain routing information necessary for forwarding the data frames.

15. The method of Claim 14 further comprising storing additionally in the content addressable memory the ones-complement of the destination addresses of the computers to which may be transmitted data frames which do not contain the routing information necessary to reach the computers; storing in a second memory means the routing information necessary to reach those computers to which data frames not containing the routing information necessary to reach 5 those computers may be transmitted; determining whether a data frame does or does not contain routing information; one-complementing the destination address of the data frames which do not contain routing 1o information; comparing the ones-complemented destination address of the data frame to the list of ones-complemented addresses in the content addressable memory; 15 returning a number indicating the place in the content addressable memory of the ones- complemented address if it is found; entering the second memory using the number as an index to retrieve the routing 20 information for the data frame, and transmitting the data frame onto the computer network which did not originate the data frame, the retrieved routing information being incorporated into the data frame. 25

30

35