CAUCHO RECICLADO DE LLANTA (CRLL)

C.4.1. O verview

In this section, we look at the address space in more detailed manner, and analyse how its m anagement is carried out on request for a new address. Note that we are looking at the addressing on a network layer or on a link layer in the first place. Releasing an unused address is another important issue of the management, and will be discussed later in Section C.4.8 Address maintenance.

The next subsection discusses addressing schemes. Subsection C.4.3 describes the address allocation operation.

C.4.2. A ddressing Schemes

A network layer address is composed o f objects called numbers. The address space, A,

includes those numbers. Addressing is to pick up numbers, concatenate them, and produce an address. A n address itself is a number and an element o f A.

There have been two addressing schemes; one is relative addressing and another is absolute addressing [DaP81] or physical addressing [Com8 8].

way. If it is hierarchical according to the network structure, an address is composed of objects, such as a network number and a local number. A local number is only unique in that network domain. A local number would be further divided into a subnetwork number and a host number. A host number is again only unique in the subnetwork domain. This type of addressing is network-relative. The Internet IP addressing [RSR8 8] is an example o f network-relative addressing.

Absolute addressing would be that an address is a unique num ber in an address space. The Ethernet addressing could be an example o f this scheme, however, it actually is produced in a hierarchical m anner — relative not to a network but to a vendor [DIX82].

The most significant characteristic of absolute addressing is that wherever a network interface is moved to, the address remains same. On the other hand, a network-relative address indicates a kind o f location of an entity; the address has to be changed when the entity moves to a different location.

Network-relative addressing has a property that each component of an address, such as network address, subnetwork addresses, and host addresses, is mapped to a specific type o f group object, such as network, subnetwork, and host. A formal notation for this is as follows:

Group(A.nets) C N e ts U { 0 }

Group(A.subnets) C SubnetsU {0} Group(A.hosts) C H ostsU G atew ays U {0}

Similarly, if any network object, net £ N e t s, has an address, the address is one of the network addresses, a £ A.nets. It is true for Subnets and H osts as well.

A ddress(N ets) C A.nets U {0} Address(Subnets) C A.subnets U {0} A ddress(H osts U Gateways) C A.h osts U {0}

Note that as far as addressing is concerned, there is no difference between gateways and hosts. They are equally treated as hosts, as each netw ork interface o f a gateway is addressed as a host.

According to a classical conception [Sho78], an address indicates where a resource is and a route tells us how to get there. A n address could be independent from a routing scheme as in absolute addressing. This is the case in a small-size network and in a broadcast network. The size o f a network may grow by interconnecting some subnetworks via store-and-forward nodes such as routers, and gateways. As the size o f a network reflects directly on the size of a routing table [Sun77], the routing is structured naturally in a hierarchical manner [K1K77]; for a large network the routing table at each node may be based on clusters at a certain level such as network, and not on the whole nodes.

To ease this scheme, hierarchical addressing, namely relative addressing may be introduced. In this case, addressing is no longer independent o f routing. It is a routing policy that decides an addressing.

If the routing hierarchy is based on the network structure, the addressing would be network- relative’, that has been traditionally the hierarchical addressing. On the other hand, if the routing hierarchy is based on a region where a united access control policy could be exercised [Cla89], an address could be region-relative. Indeed, an addressable entity, eg. a region or a network, is a unit o f access control on the network level. The addressability provides basic reachability. Accessibility is the result o f the access control over the reachable entities. Therefore, the practical reachability is derived by accessibility.

In the further discussion, we only look at the network-relative addressing as an example. Absolute addressing could be seen as identical to intra-domain addressing o f network-relative

addressing.

C.4.3. Address allocation

The allocation o f an address includes the following operations:

i. learning the range o f address space from which one can assign an address for an object in an environment

ii. selecting an address from the free address space for a group object.

1. Network Address Allocation:

i. learn a Network Class o f a Network object, g £ Requesting at the current time, U, which needs an address

ii. learn the Network Number range for a given Network Class (The Internet IP uses an 32-bit address for hosts. As it is network- relative, a network part is the first 8 or 16 or 24 bits for class A or B or C network address [RSR88])

iii. choose a Network Number from the part of the range which has not been assigned to anything yet

iv. produce a Network Address by concatenating the Network Class and the Network Number.

2. Subnetwork Address Allocation:

i. learn the Subnetwork Number range for a given Network A d ­ dress

ii. choose a Subnetwork Number from the part o f the range which has not been assigned to anything yet

iii. produce a Subnetwork Address by concatenating the Network address and the Subnetwork Number.

3. H ost Address Allocation:

i. learn the Host Number range for a given Subnetwork Address ii. choose a Host Number from the part o f the range which has not

been assigned to anything yet

iii. produce a H ost Address by concatenating the Subnetwork ad­ dress and the Host Number.

We formalise the host address allocation in the following. We define new fu n ctio n s to specify various ranges;

NumberSpace{x £ A) = Y C A

where Member(g £ the range o f G rou p(Y )) C Member o Group(x) Host Number Space = NumberSpace \ A.subnets

NumberSpace \ A.subnets means that the dom ain o f fun ction is restricted to the address space for subnets in A, A.subnets.

Host address allocation, for instance, can be form alised as follows: i. 3s £ A.subnets for a given group object, g £ Requestingti ii. the Host Number range is H ostNum berSpace(s) C A

iii. choose n £ Concatenate(s, H o s tN umber Space(s)) — Assignedt; where Concatenate is a function, Concatenate: Ax A—► A to produce an address from two numbers.

iv. a host address, a £ A is obtained by a = Concatenate(s, n).

A new address can be taken, if and only if a group object has not had an address yet. A new number is taken in (ii) above, if and only if there is a free number in H o s tN umber Space(s) to be assigned to this type of group object. This condition guarantees the address to be unique in the address space. The formal notation for this is as follows:

3a £ A for a given g £ H o stsn Requestingti if and only if

a A ssignedt i \ and

C oncatenate^, Host Number Space(s)) — Assignedti {0}