RIPE DRAFT

19 Apr 1993

      Dear All,

Below is a first draft of the "Hints" supporting 
documentation. The production of this document is 
a minuted action item from the last RIPE meeting.

The content of the document and the questions below
will be discussed at the RIPE meeting next week.  Please
bring your comments to the meeting.

1.  Class D procedure - is the assignment of these within
     the scope of this procedure?

2.  The issue of non-contigous subnets (eg multihomed orgs 
    using a subnetted Class B) and the potential difficulties
    thereof? do we wish to give advice on this

2.  Is there a need for a short Appendix describing how to find
    a NOC of Last resort (cf App 2 on service providers)?

----------------------cut here----------------------

       DRAFT  DRAFT  DRAFT  DRAFT DRAFT  DRAFT  DRAFT
              DRAFT DRAFT  DRAFT  DRAFT  DRAFT

       HINTS FOR ORGANISATIONS REQUESTING IP NETWORK
                          NUMBERS

                     Bob Day, Anne Lord

                         ripe-draft

This document is intended  to  complement  and  support  the
information  described  in  the  "European IP network number
application form and template" (see RIPE document ID:  ripe-
83).    The  aim  of  the  document  is to guide you in your
choice of class of IP network number so that you choose that
which is best suited to your organisations needs.

The  document  is  motivated  by   the   large   number   of
applications  that are received for Class B address that are
not in fact allocated. This accounts for  approximately  90%
of  all  class  B  applications.  It is a time consuming and
often lengthy process explaining to organisations why  their
application has been rejected, or why it is taking longer to
process, which we hope can be lessened with the  publication
of this document.
                           - 2 -

Contents

1     Background ................................................

2     IP network number scarcity ................................

3     IP addressing .............................................

3.1   Subnetting ................................................

4     Choosing the Class of Network Number ......................

4.1   Using a Single Class C Network Number .....................

4.2   Using a Block of Class C Network Numbers ..................

4.3   Applying for a Class B Network Number .....................

Appendix 1: Supernetting ........................................

Appendix 2: What to do if you need a Service Provider ...........
                           - 3 -

Copyright c 1993

Whilst every effort has been  taken  to   ensure   accuracy,
the   RIPE NCC does  not  accept any responsibility for loss
or damage arising from the use of information  found  within
this document.

Material from this document may be  incorporated  in   other
technical documentation,  subject  to  prior agreement from,
and acknowledgement of, the RIPE NCC.

1    Background

The  arrangements  for  the  allocation  of  Internet   (IP)
network   numbers have  recently  been  revised.  Previously
these numbers were assigned only by the Network  Information
Centre  (NIC) of the Defense Data  Net- work  (DDN)  in  the
US.  This was done by  consensus  on  behalf  of  the  whole
Internet  community.  Following the change of  arrangements,
the  DDN  NIC  still  has  overall  responsibility  for  the
allocation  of  network  numbers  but it has  delegated  the
actual  assignment  process  on  a regional basis.

In  Europe  the   delegated   authority   is   the   Network
Coordination    Centre  (NCC)   run   by   RIPE   under  the
auspices of RARE.  The NCC has further delegated a number of
IP  ``service  providers''  to assign  numbers  for networks
connecting to their respective service networks.  The <local
example>  is one of these service providers (it provides the
<local cc: IP Service)  and  consequently  now  handles  the
assignment  of "C" network numbers to networks connecting to
the <network>.

2    IP network number global scarcity

The Internet authorities are  increasingly  concerned  about
the   possi- bility of exhaustion of the IP address space as
a result of the recent explosive  growth  of  the  Internet.
They   have   decided   upon  certain measures to attempt to
conserve address space, and other  solutions  are  currently
under debate in the community.  This is now a matter of some
concern.  Further detail on the  measures  decided  upon  so
far is given in Appendix 1 of this document.

One of the measures  currently  practised  by  the  Internet
community  is to carefully review each and every application
for network numbers with respect to its merit  on  technical
grounds.   Strict criteria are applied to all organisations,
regardless of type, to ensure  that  the  remaining  address
space is distributed as effectively as possible.
                           - 4 -

3    IP Addressing

The IP address of an end system attached to an IP network is
composed of two elements:

- the network number identifying to which  network  the  end
system  is  attached  (uniquely  amongst all the IP networks
that constitute the Internet);

- the  host  number  identifying  the  end  system  on  that
network.

The entire address is a 32 bit quantity. The usual means  of
represent-  ing  an  address  is  to write it as a series of
four decimal numbers,  each  representing  8  bits  of   the
entire   address,   and   separated   by periods.  Thus, for
example, the address:

        192.100.100.27

would represent the end system numbered  ``27''  on  the  IP
network    with   number    ``192.100.100''.     It  is  the
requirement for global uniqueness of the network number that
leads  to  the  need for co-ordination in  the assignment of
these numbers.

IP network numbers are divided into a number of ``classes'',
each   of  which   allows  a different maximum number of end
systems to be attached to  the  network  it  represents  (ie
gives   a   different  maximum   number   of  possible  host
addresses).  Of these there are two  classes  that  will  be
relevant  to  an  organisation applying for a network number
through  the <local sp/nic>.  A ``Class C''  network  number
will  allow  the  attachment  of  up  to  256 end systems, a
``Class B'' network will allow up  to  65,636  end  systems.
(In each case two of the end system numbers are reserved for
conventional uses, meaning that the number of  host  numbers
available in practice is 254 or 65,634 respectively.)

These figures come about because a Class  C  network  number
always  occupies  the  first 24 bits of the full IP address,
leaving  8  bits  for  the  host  number.   This  gives  the
possibility  of  256 different host numbers, of which one is
reserved as a conventional ``broadcast'' address.  A Class B
network  number  only occupies the first 16 bits of the full
IP address, leaving 16 bits for  the  host  number.   An  IP
implementation  can  determine the class of a network number
by examining the first two bits.  If only the first of these
is set - ie the top byte is in the range 128 - 191 - it is a
Class B number.  If both bits are set (and the next  bit  is
unset) - ie the top byte is in the range 192 - 223 - it is a
Class C number.
                           - 5 -

Recently there has been growing interest in the use of Class
D  numbers  as  well.  These are used to create IP multicast
addresses - ie if  a  system  transmits  a  datagram  to  an
address  within  a  Class  D  network  it  will be delivered
simulataneously to a group of hosts, rather than to a single
host.  IP  multicasting  has  applications  in  the  area of
coperative   working   and   conferencing,   as   well    as
(potentially) in the support of routing protocols. A Class D
network number has the top three bits set - ie the top  byte
has the value 224 or greater.

3.1 Subnetting

Associated with each IP  address  is  an  ``address  mask''.
This  is a 32 bit quantity that marks, in a bitwise fashion,
which bits of the address are to be treated as  the  network
number  component  and  which  are to be treated as the host
number component.  Where a bit is set in the  address  mask,
the  corresponding  bit  of  the address is considered to be
part of the network number field.  Where the bit is unset in
the  address mask, the corresponding bit is considered to be
part of the host number field.

For  a  Class  C  address  the  default  address   mask   is
255.255.255.0  (ie  the  top  24  bits  contain  the network
number).  For a Class B address the default address mask  is
255.255.0.0.

By use of a non-default address mask, it is possible for the
administrator  of  a Class B network number to break it down
into a number of Class C ``subnets''.  These could then, for
example, be assigned one per department in a University, and
routers could be used to connect these together.  This would
allow  a  site  network  to  be  broken  down  into a set of
autonomous networks, whilst the network as a  whole  appears
to the outside world to have a single (Class B) number.

As an illustration, assume that an institution has the Class
B  number  128.100  assigned to it.  The administrator could
create 256 Class  C  subnets  by  specifying  a  non-default
address  mask of 255.255.255.0.  This would allocate the top
8 bits of the host number field to be an  extension  of  the
network  number  field.   Hence  the  set of Class C numbers
128.100.0 - 128.100.255 would become available.   Of  these,
the  first and last in the range should not be used, as they
have conventional meanings.  This  would  leave  up  to  254
Class C numbers for use.

In principle  subnetting  need  not  be  done  on  an  8-bit
boundary  eg  an address mask of 255.255.240.0 could be used
to produce 16 subnets (14 of them useable), each with a  12-
bit host field.  In practice, however, subnetting is usually
confined to an 8-bit boundary.
                           - 6 -

Subnetting is  thus  a  technique  of  moving  the  boundary
between  the  host  and  network number parts of an address.
For it to be useful,  the  IP  implementations  of  all  end
systems  on  the network involved must support it.  All must
also use the same, centrally defined address mask.

4    Choosing the Class of Network Number

An organisation that requires more address space than  would
be  provided  by  a  single Class C  network number  will by
default  receive a group of Class C numbers  instead.   This
implies  that  it  will need to structure its  site  network
into separate,  interconnected  Class  C  networks.

The rest of this section goes into  detail  as  to  how  the
decision as to which class of address to apply for should be
approached.  The aspects to be considered when  making  this
decision are as follows:

- the current requirement
  in terms of the the number of end systems to be connected;

- the likely expansion over the next one or two years;

- the feasibility or otherwise of routing between networks
  on site, if multiple Class C networks are to be used.

4.1    Using a Single Class C Network Number

If the requirement in terms of end systems to  be  connected
are   modest  -  perhaps   a   few   tens  of  systems to be
connected (max 255 hosts) - a single Class C network  number
will  be  sufficient.   This  is   the   simplest  and  most
trouble-free, situation.

4.2    Using a Block of Class C Network Numbers

If it is likely that there will be a few hundred end systems
connected  over   the  next  year  or two the default choice
will be to ask for an assignment  of  a  block  of  Class  C
network   numbers.    These  will  need   to  be   organised
internally as a set of interconnected networks, using an  IP
router  (or routers) as the  means  of  interconnection.   A
common organisation  is for the site's network  operator  to
assign  one  Class C network per department, and to  connect
these  together  via  a  site ``backbone''.   For   example,
assume that the site has been allocated four Class C network
numbers:  192.100.100  -  192.100.103.   These    could   be
assigned  to three different departments and a backbone, and
a sin- gle router used to interconnect  them,  as  shown  in
Figure 1.
                           - 7 -

                192.100.100 (backbone)
===o==============o===============o============o===
   |              |               |            |
 +---+          +---+           +---+        +---+     Connection
 | r |          | r |           | r |        | r | --> to service
 +---+          +---+           +---+        +---+     provider or
   |              |               |                    other
===o========   ===o=========   ===o========
 192.100.101    192.100.102     192.100.103
  (Dept A)       (Dept B)        (Dept C)

Figure 1: Interconnection of Class C Networks via a Backbone
Network

Alternatively, the four networks  might  be  connected   via
a   single  router,   as  shown  in Figure 2.  The choice of
interconnection method will be dictated by the conditions on
site,  but  in  all  cases some form of IP routing equipment
will be needed.

          192.100.100 (Dept A)       +---+
         ============================|   |
                                     |   |
          192.100.101 (Dept B)       | r |
         ============================| o |
                                     | u |    Connection
          192.100.102 (Dept C)       | t |--> to service
         ============================| e |    provider or
                                     | r |    other
          192.100.103 (Dept D)       |   |
         ============================|   |
                                     +---+

Figure 2: Interconnection of Class  C  Networks  via  Single
Router

A consequence of the recent rapid growth of the Internet  is
that   the  number  of  network  numbers  that  have  to  be
configured into regional and international routers has  also
grown  rapidly.   This  means  that  these routers'  routing
tables  have also grown to the point where there is  concern
as  to  whether  they will continue to operate  efficiently.
To combat  this  problem  the concept of ``supernetting'' is
being   intro-  duced.   This  is  outlined  in  Appendix  1
(although it is  not  necessary to understand the concept to
apply  for  a  network  number).  A practical consequence of
this move is that a request for multiple Class  C  net- work
numbers  will always result in a contiguous block of numbers
                           - 8 -

being assigned, and that the size of the block  will  always
be  a  power  of  two  (ie 2, 4, 8, 16 or 32 network numbers
etc).

4.3 Applying for a Class B Network Number

There may be some circumstances where the use  of  a  single
Class  B  network  number,  rather  than  a block of Class C
numbers is justified.  This may be because the number of end
systems  to  be  connected  is  so  large  that  it  becomes
cumbersome to use a block of Class C numbers.  The guideline
given  by  the  Internet  NIC  (in  RFC 1366) is that a site
network should utilise a Class B number if, based  on  a  24
month projection, it requires:

- more than 32 network numbers (or subnets), AND

- it has more than 4096 end systems to connect.

The Class B  network  number  could  then  be  subnetted  if
necessary, according to the site requirements.

Site networks  that  anticipate  requiring  less  than  this
amount  of address space should, under normal circumstances,
apply for a block of Class C network numbers.

Another potential reason for the use of a  Class  B  network
number  is  that it may be infeasible for the institution to
do the IP routing required on its site network if a block of
Class C numbers is used.  As shown in Figures 1 and 2 above,
this will require the installation of  routing  equipment  -
either  purpose-built  routers  or end systems equipped with
multiple LAN interfaces and IP routing software.  This might
be  impractical  in  some  cases, on the grounds of existing
investment in equipment.  It might also be impractical in  a
situation  where  the site network is multi-protocol and the
routers cannot handle all the protocols involved.  MAC level
bridging might then be required, along with a single network
number across the entire network.

In making the decision as to whether a  Class  B  number  is
necessary,  note  that many purpose-built routers can bridge
as well as route (so-called  ``brouters''),  so  it  may  be
possible  to route IP whilst bridging other protocols.  Note
also that  the  ``supernetting''  development  described  in
Appendix  1  means  in  theory that the use of IP routers on
site can be avoided in the case where a  suitable  block  of
Class C network numbers has been assigned.

To help the NICs  involved  determine  whether  there  is  a
sufficient   case   for   a  Class  B  network  number,  the
organisation is asked on the ``European  IP  network  number
                           - 9 -

application  form''  to  supply  information relating to the
number of hosts and the number of subnets, in  use  now  and
predicted  for  one  and  for two years' time. Besides there
being a sufficient number of hosts to address, the NICs must
determine  that  the  network  cannot  be engineered using a
number of contiguous  class  C  networks.   If  the  network
consists  of  a  large  number  of  physical  networks  with
relatively small numbers  of  hosts  on  each,  it  will  be
necessary  to consider subnetting class C networks.  A large
number of subnetworks alone is not sufficient  justification
for  allocation  of  a  class  B  address.  The guideline in
RFC 1366 will be applied rigorously.

The procedure for deciding whether a Class B number  can  be
allocated  is  first  that  the  <nic/sp>  will  assess  the
case  and,  if  it agrees,  will   recommend  to  the   RIPE
NCC  that  a  Class  B  network  number  is allocated to the
organisation concerned.  The RIPE NCC will also  review  the
case  briefly  and  make a decision in consultation with the
<nic/sp> and the organisation concerned.   Because  of  this
two  stage  consultation  process  the application will most
likely  take longer than normal to be dealt with.
                           - 10 -

                          Appendix 1

                         Supernetting

One of  the  perceived  problems  arising  from  the   rapid
growth   of  the Internet  is  the  consequent growth in the
size of the routing tables held in the various regional  and
international   routers.   The  increased  pressure  to  use
multiple Class C network numbers, rather than a single Class
B  number,  in  order to economise on the use of the  latter
class will add to the size of these routing tables.

As a way of mitigating this problem it has been  decided  to
use  a  route  aggregation   scheme  colloquially  known  as
``supernetting''.  (It is also known  as  CIDR  -  Classless
Inter  Domain  Routing,  and  is  described in detail in RFC
1338.)

The key to the scheme is that  where  a  block  of  Class  C
network  numbers is assigned to an organisation's network it
is done so as a contiguous block of a size that is  a  power
of  two.   This  means  that  for  routing purposes  it will
then be possible to treat the entire block  as  a  sin-  gle
network, albeit with a special address mask.   (The  address
mask associated  with  an  IP address is a 32  bit  quantity
that  marks, in a bitwise fashion, which bits of the address
are to be  treated  as  the network  number  component   and
which  are  to be treated as the host number component.  For
a  Class  C  address   the   default   address    mask    is
255.255.255.0   -  ie  the  top  24 bits contain the network
number.  For a Class B address the default address  mask  is
255.255.0.0.)

To illustrate this, take as an example  the  block  of  four
Class  C  net- work numbers 192.100.100 - 192.100.103.  This
can be treated as a sin- gle network number  192.100.100  by
using  an  address  mask that specifies the  network  number
component to be only 22 bits rather than 24 bits.   This  is
shown in Figure 3.

           <--------network-------><---host-->
          +--------+--------+--------+--------+
          |  192   |  100   | 100    |        |
          +--------+--------+--------+--------+
 address   11111111.11111111.11111100.00000000
    mask                                    (ie. 255.255.252.0)

Figure 3: Illustration of a Supernetting Address Mask
                           - 11 -

Because the block of network numbers is of size  four,   and
has   been  assigned   to   start  with a value divisible by
four, it is certain that the bottom two bits of  the  normal
24  bits  used  for a Class  C  network number will be zero.
Therefore the address mask can be set to make it appear that
these  two  bits are part of the host  number  component  of
the address, and consequently  that  the  networks  numbered
192.100.101   -  192.100.103  are  subnets  of  the  network
numbered 192.100.100.

The technique is called ``supernetting'' because it  employs
a   similar  principle   to  the  established  technique  of
``subnetting''.  In the  latter  case  bits  from  the  host
number  component  of  an  address  are  made  part  of  the
network number component, in  effect  creating  a  range  of
subnets  from  a  single  network  number.   It will work in
theory for  any size  block  of  network  numbers,  provided
the  block  is contiguous and the ``power of two'' criterion
is satisfied.

Supernetting  can  work  in  practice  only   if   the    IP
implementations   of  all  equipment  handling  it have been
modified to understand it.  Other- wise the special  address
mask involved will appear  invalid,  and  the implementation
will  treat  each   network   number   in   the   block   as
representing  an  individual  network.   Hence  if  all  the
routers  in  a regional  network  to  which the organisation
is  attached  do  implement supernetting they will treat the
entire   block   as   representing   a    single    network.
Consequently, in this example, there would be only one entry
in the regional routers' tables rather  than  four,  but  IP
traffic  for any network contained in this block would still
be routed correctly to the organisation concerned.

Depending on implementation of  supernetting  by  the  major
router   ven-  dors,   it   is  expected  that  regional and
international routers will adopt this scheme in near future.
Follow the recommendations of the provider involved.

If  all  end  systems  on  the  network  of  a    connecting
organisation,   and  the   router   used  to  connect to the
outside world implement supernet- ting it will  be  possible
to  construct  the  network   using   a   block  of Class  C
numbers  and  without the need for router(s) internal to the
                           - 12 -

network.   However, it seems very unlikely that this will be
the  case in  the  immediate  future,  and  it  is  best  to
assume  that traditional routing techniques will be required
within the site.
                           - 13 -

                        Appendix 2

              What to do if you need a Service Provider

If your organisation is planning to connect to the  Internet
in  the near future, then it is recommended that you do this
via an IP Service Provider.  If  you  are  unsure  who  your
service  provider  would  naturally  be, then you can fax or
telephone the  RIPE  NCC  who  will  send  details  of  your
connectivity   requirements to a mailing list maintained for
this purpose.  Please supply your contact information  which
individual  IP providers who have subscribed to the list can
use to contact you. If you are sending a  fax,  please  mark
it:

For the attention of :   ip-provs@ripe.net

We will then  transcribe  your  details  to  our  electronic
mailing  list.   Note  that  this  is  the extent of the NCC
involvement  -  it  is  a  matter  for  individual   service
providers to decide whether to follow up such a request.

RIPE Network Coordination Centre        tel: +31 20 592 5065
Kruislaan 409                           fax: +31 20 592 5090
1098 SJ Amsterdam                       email: hostmaster@ripe.net