M.I.T. Lab. for Computer Science Network Implementation Note No. 5

                                                          February 15, 1979

                               LCS Net Address Format

                                    Noel Chiappa
                                     David Clark
                                     David Reed

          This note defines the current format  of  addresses  in  the  LCS
          Local Net and their translation into LCS hardware headers as well
          as InterNet addresses and CHAOS Net addresses.

          This  note  is an informal working paper of the M.I.T. Laboratory
          for Computer Science.  It should not be  reproduced  without  the
          author's  permission,  and  it  should  not  be  cited  in  other

                                  February 15, 1979

               This note defines the current format of addresses in the LCS
          Local Net and their translation into LCSNet hardware  headers  as
          well  as  CHAOS  addresses  and  InterNet  addresses. Address and
          protocol numbers specific  to  the  LCSNet  are  also  temporaily
          defined in this note.

               Note  that  the proposal stated in this is not final, but is
          motivated by other considerations. The LNI as currently  designed
          contains sophisticated name matching facilities of unknown worth.
          It  is  not yet clear whether this hardware is useful or not, but
          we felt that we should at least explore  the  possibilities,  and
          have  thus designed the protocol to take maximum advantage of the
          hardware's features. Depending on the results of our  preliminary
          work  with the name matching hardware, we may or may not keep< it
          in future versions of the LNI.

               The basic hardware format of a LCS-LN packet is as follows:

          DPNM    31                               0

          DPN     31                               0

          OPN     31                               0

          LEN     15               0

          DATA    8n-1             0

               (Unless others stated, all bit numbers are  in  decimal  and
          all others are in octal.)

               The  DPNM,  DPN,  OPN  and LEN fields are all defined by the
          hardware. The DATA field contains LEN bytes  of  data.  The  DPNM
          (Destination   Process   Name   Mask)   together   with  the  DPN
          (Destination Process Name) specify the destination  address,  and
          the OPN (Originating  Process Name) is the sender's address. This
          field  is  actually  unused  by the hardware, but is reserved for
          future use.  The workings of the DPNM and DPN are as follows:  in
          every  LNI  there is an array of sofware loadable names (the Name
          Table, NT), each with  a  DPNM  and  a  DPN.  The  name  matching
          algorithm is a follows: for each bit in the name, if the two bits
          in  the  DPN's  are  equivalent or if either DPNM bit is on, then

                                  February 15, 1979

          that bit matched. If all the bits matched, then the name matched.

               The basic format of the headers (imposed by the sotware)  is
          as follows:

          DPNM    31 Type  23                       0

          DPN     31  Rsd  23                       0

          OPN     31           Reserved            0

          LEN     15               0

          DATA    8n-1             0

          "Rsd"  stands  for  "Reserved" and "Type" is the Local Type. As a
          general rule the value of 0 for any field is reserved for  fixing
          catastrophes   not   yet   visible  to  the  eye.  The  following
          conventions are to be used in assigning  type  numbers:  types  1
          through  077 have a common address format, specified below, types
          100 through 277 are reserved,  and  types  301  through  377  are
          available  for (experimental) protocols which may use the next 24
          bits of address as they see fit.

               The intention of the above is that bridges be able to easily
          recognize, via the top bit(s) whether or not  the  address  field
          fits  a  common format. Clearly there are enormous advantages, in
          the building of bridges and other areas, where it is desirable to
          have a common address format, but it is obviously also  desirable
          not   to  hamstring  the  development  of  other  very  different
          protocols (such as a distributed process system or a  distributed
          file system) which would fit poorly into the classical addressing

               For  types  that  use the common address format, the address
          format is:

                   00  Type    SNet      Rsd     Host
                  31 29     23       15       7        0

                                  February 15, 1979

          The field marked "Rsd" is reserved for future use. It is intended
          that it be used for address expansion should it become necessary,
          but it may be used for whatever appears  appropiate,  such  as  a
          "Flag" field.

               SubNet  numbers  and Host Numbers (per SubNet) are currently
          being assigned in common with  the  AI  Lab's  CHAOS  Net  in  an
          attempt  to  maintain  an  MIT  wide  addressing  unity. The main
          repository  for   these   numbers   is   in   the   online   file
          AI:SYSENG;HOSTS > on MIT-AI.  numbers specific to our net are:

          Number  SubNet

            0             Reserved
            1-7           Unassigned
            10            LCS LN
            11-377        Unassigned

          Subnet 10
          Number          Host

            0             Reserved
            1 - 7         Unassigned
           10             CSR 11/40
           11             DSSR 11/70
           12 - 17        Unassigned
           20             VAX
           21 - 377       Unassigned

          It is not intended that this note be the official source for host
          and  subnet  numbers;  rather,  there  will be an online database
          which at the moment is the specified ITS file.

               The assigned protocol types are:

          Type            Use

            0             Reserved
            1             InterNet Datagram (Internal)
            2             CHAOS Packet
            3-77          Unassigned
            100-277       Reserved
            300           Reserved
            301           InterNet Datagram (Outbound)
            302-377       Unassigned

                                  February 15, 1979

               In all cases the specified header is followed immediately by
          a packet of the specified type  as  defined  in  the  appropriate
          reference.  [1,2]  At the moment, the format of outbound InterNet
          packets is as follows:

                   11   000         Rsd       Ext Net
                  31 29     23               7        0

          where "Ext Net" is the major net number of the  destination.   In
          internal  InterNet  packets, the mapping from InterNet Headers to
          MIT addresses is simple; the 24  bits  of  "Address"  after  "Net
          Number"  in  the  InterNet  header  have  the  same format as the
          adddress of a  common  address  format  packet,  and  are  copied
          directly into the header. In the future, some attempt may be made
          to support variable address format addresses on inbound packets.

                                  February 15, 1979


          [1]  Postel,  Jonathan  B., "InterNetwork Protocol Specification,
          Version  IV,"  Information  Sciences  Institute,  University   of
          Southern California, IEN 54, September 1978.

          [2]  Greenblatt,  Richard  G.,  and  Moon, David, "CHAOS Protocol
          Specification," Artificial Intelligence Laboratory, M.I.T., 1978.