IP addressing, and gaining IP's

From: wally%net1.ucsd.edu@flash.bellcore.com (Wally Linstruth) (tty00)

        To: tcpgroup

        Regarding: IP addressing

         The  intent  of  this paper is to  document  the  backgroun behind the current IP address assignments which I have offered to  coordinate.   The proposed scheme has been reviewed by Phil Karn,Bdale Garbee and (verbally with) Mike Chepponis, all of whom have encouraged that it be used.

             Phil's  code  does  NOT  currently  support  the  subnetwork 

        aspects of the scheme but will do so in the future.   There is no 

        real  reason  for  any national coordination of  these  addresses 

        until  actual  networks or at  least  geographically  coordinated

        groups of experimenters are formed.

             I  have offered to issue and keep track of SUBNET  addresses

        and  their  "owners"  who are  presumably  responsible  *NETWORK*

        implementors and managers.

 The  basic premise behind the proposed plan is that  amateur

 radio  networks will be politically defined.   The plan is  based 

 upon  the  presumption  that current voice networks  serve  as  a

 proper  analog by which to predict general characteristics of the

  as yet unconstructed digital networks.   Political entities  will

 build networks; funded, controlled, maintained and used primarily

        by their own members and guests.

Each  of these separately managed networks should be  vieweas  a  subnetwork  of AMPRNET (with the  idea  being  to  somehow

rationally  partition the 044.xxx.xxx.xxx AMPRNET address space). Each  subnetwork within AMPRNET will maintain routing tables  for  its  own  constituents.   Each will provide its own hosts  (TACs,Gateways, i.e. the mechanism by which users with simple terminals and AX25 level 2 boxes will access network resources),  switches,rules  (network  administration),  security  measures  and  quite possibly its own link level protocols.

The  natural  limitations on span of control  will  probably limit  the  service  area of each of  these  networks.   This  is another factor leading to the partitioning of the AMPRNET address space with respect to separate subnetworks.

  This  partitioning  of  the  address space  will  allow  for much  simplified  routing tables in each  host.   Internetworking gateways will connect these independently controlled subnetworks.Each  gateway will maintain routing tables only for  local  hosts and for gateways to other networks.  Hosts and  relay switches on a   given  subnet  will  need  to  maintain  routing  information regarding  only  members  of that subnet and  gateways  to  other networks.   The  required routing tables should prove to be  very  manageable  and make any kind of geographically  based  hueristic addressing schemes such as ZIP codes, area codes etc. moot.  

1  I  would  also  like to propose that we  coordinate  logical     network  names and their corresponding addresses based  on  these political   network  subdivisions.    The  concept  of  a  naming 

 convention  which maps directly into an IP address is purely  for

        the  convenience  of  network developers and  is  not  considered necessary.  There is,  however, some good reasoning behind making network and host names hierarchical and meaningful to end  users. It  will  considerably aid in bootstrapping the initial  networks and in being comprehensible to the non-network folks who will  be  the  primary  users  of these networks.   The  naming  convention proposed   is  of  the   form   USERID@HOST.SUBNET[.AMPRNET.RES]. 

  

WESTNET,  SBARCnet  (Santa  Barbara ARC) and  GFRN-net  represent

        three  hypothetical  networks  with which this  writer  could  be

        involved, perhaps as a provider of gateway and/or host services.

Each  of  these  subnetwork entities could have  a  distinct      address  and  perhaps several internally  administered  

host/user  addresses.

             [NOTE:  Throughout this paper,  Host or Host/User represents

             any  host  or any user running IP protocols that has  direct

             network  access.  Also,  for the purposes of  the  following

             example,   WA6JPR  is  not  a  network  address,  rather  it

             represents  a user-id on a local host.   It is the  writer's

             opinion that the majority of packet users for the forseeable

             future  will  be using simple TNCs connected  to  hosts  via 

             AX.25 level 2 protocols.]

 WA6JPR  may  be "a user" on hosts on more than  one  network      such  that  a station in Washington D.C.,logged onto  an  AMPRNET       host,    may    send    internet    traffic    successfully    to       WA6JPR@JPRHOST.WESTNET  (this traffic would be routed to  Westnet

via various AMPRNET gateways and subnetwork level relays and then  to  a  Santa  Barbara  host known internally  by  Westnet  to  be

        reachable  via  the  W6AMT-2  switch).   Traffic  could  also  be  directed  to  Wally@SBARC   (presuming  that  the  Santa  Barbara Amateur  Radio Club maintains a message server host gatewayed  to the AMPRNET catenet).

Based  upon  the  presumption  of  the   AMPRNET/SUBNET/HOST  hierarchy,  it  would  seem  that we could easily decide  how  to allocate  the 044.xxx.xxx.xxx 24 bit IP address field  such  that  there  are bits allocated for a sufficient number of individually managed  subnetworks  while leaving  a  correspondingly  adequate number  of  assignable bits for the internal addressing needs  of   each individual subnetwork.

Accordingly,   the  following  is  proposed  as  an  initial  addressing  scheme and methodology for address assignment.   [Bit numbering is per RFC-960 Pg.2]

                                        2

        Bit  8  to  be 0 for USA stations and  1  for  non-USA  stations. 

        [Note.   This  is  not  meant  to imply a  geographic  basis  for

        assignments.   It  is  meant to provide a very  quick  means  for 

        segregating FCC controlled participants from non-FCC stations.]

        Bits  9 - 18 to represent politically separate subnetworks within

        AMPRNET.   These  bits  are to be assigned in an  inverse  binary

        sequence   (see   example   below)  beginning  with   the   *MOST

        SIGNIFICANT* bit first.

        Bits 19 - 23 to be unassigned and reserved for future  allocation

        as  network addresses,  to network administrations for internally

        assigned  host  and/or user addresses,  to a combination  of  the 

        above or to a completely new intermediate class of addresses.

        Bits  24  - 31  to be used within  politically  separate  AMPRNET

        subnetworks for individual hosts,  switches, workstations etc. as

        determined  by  local  network  administration.    It  would   be 

        recommended  that these bits be assigned in binary sequence  with

        the *LEAST SIGNIFICANT* bits being assigned first.

             The  resulting  network  addresses would be as follows:

        AMPRNET

        ||

        || SUBNET----+

        || |         |

        || |         | HOST--+

        || |         | |     |

        44:0...127:000:0...255------- 32,768 addresses assignable

        44:0...127:001:0...255--+ | +- 1,015,808 addresses reserved

        44:0...127:031:0...255--+

        44:0...127:032:0...255------- 32,768 addresses assignable

        44:0...127:033:0...255--+ | +- 1,015,808 addresses reserved

        44:0...127:063:0...255--+

        44:0...127:064:0...255------- 32,768 addresses assignable

        44:0...127:065:0...255--+  | +- 1,015,808 addresses reserved

        44:0...127:095:0...255--+

        44:0...127:096:0...255------- 32,768 addresses assignable

        44:0...127:097:0...255--+ | +- 1,015,808 addresses reserved

        44:0...127:127:0...255--+

        44:0...127:128:0...255------- 32,768 addresses assignable

        44:0...127:129:0...255--+ |  +- 1,015,808 addresses reserved

        44:0...127:159:0...255--+

        44:0...127:160:0...255------- 32,768 addresses assignable

        44:0...127:161:0...255--+ | +- 1,015,808 addresses reserved

        44:0...127:191:0...255--+

        44:0...127:192:0...255------- 32,768 addresses assignable

                                        3

        44:0...127:193:0...255--+ |  +- 1,015,808 addresses reserved

        44:0...127:223:0...255--+

        44:0...127:224:0...255------- 32,768 addresses assignable

        44:0...127:225:0...255--+ |  +- 1,015,808 addresses reserved

        44:0...127:255:0...255--+

        44:128:xxx:xxx----------+ | +- 8,388,608 addresses assignable (non USA)

        44:255:xxx:xxx----------+

   The  above  allocation and assignment scheme allows  network (subnet)  and  intranet  (host/user) addresses  to  begin  to  b 

 immediately assigned to experimenters while retaining the largest

        possible  contiguous  block of unassigned bits whose  assignments can  be  defined  in  the future with  little  or  no  impact  on previously   allocated   addresses.    The  USER  @  HOSTNAME   .   SUBNET/ADMINISTRATION  naming scheme represents a  human-friendly 

        network  naming  convention  which  maps  easily  into  numerical   network  addresses.   I  believe that the above  approach  is  in

 general  conformance with the requirements of RFC-950,  "Internet  Standard Subnetting Procedure."

The  numbering scheme as initially proposed allows for up to

        1024  AMPRNET  subnetworks  of up to 256 hosts in the  USA  while  retaining  five  bits  for  future  expansion.    That's  262,144

  individual AMPRNET addressable entities.   If the proposed method

of  address  assignment is followed and we run out  of  Host/User

        addresses before we run out of network addresses,  we can  simply  pick  up  the  least  significant reserved bit  and  assign  more Host/User addresses.   Conversely,  if network addresses are more popular  we  could easily expand by taking the  most  significant reserved  bit and allocating it for network  addressing.

        If it should become clear that every user on a network needs  his or her own IP address, each network could allocate user blocks in 256  user  increments from the least significant  reserved  bits.  

Possible  combinations  are  1024 networks each with up  to  8192

individually  addressable units or 2048 networks each  with  4096

 hosts/users (8,388,608 individually addressable entities).

The  writer presumes that 8 million plus addresses ought  to last the US amateur population for some time to come. All we needto  do to avoid painting ourselves in a corner is to assign  them  in a logical sequence rather than randomly.

                                        4

 The  following table serves as an example of the  "high  bit first"  network  address  assignment table and  some  actual  and   requested initial networking assignments.

"this"            44.000.000.xxx      ;special case

KARNnet     44.064.000.xxx      ;network admin: KA9Q

BDALEnet   44.032.000.xxx      ;network admin: N3EUA

DCnet1         44.096.000.xxx      ;network admin: WB6RQN

SOCALnet1 44.016.000.xxx      ;network admin: WB5EKU

DCnet2         44.080.000.xxx      ;network admin: WB6RQN

SOCALnet2 44.048.000.xxx      ;network admin: WA6JPR

PITTNET     44.112.000.xxx      ;network admin: N3CVL

   next           44.008.000.xxx

   next           44.072.000.xxx                   

   last            44.063.000.xxx

  "all"           44.127.000.xxx       ;special case

 5