Transcript Chapter 4

Classful Internet Addresses
Chapter 4
Universal Identifiers
• Designers of TCP/IP determined that each host on
the internet would have a 32-bit identifier called
an IP address
– The address of the network is included in the address
(the prefix) of the host on the network
– The IP addresses of hosts on a network share a common
prefix
• Each address is a pair (netid, hostid), where netid
identifies the network, and hostid identifies the
host on that network
The Original Addressing Scheme
• The original addressing scheme was called
classful
• See Figure 4.1
– Class A addresses - used for few networks with many
hosts
– Class B addresses - used for intermediate size networks
– Class C addresses - used for networks with < 28 hosts
– Class D addresses - internet multicasting
– Class E addresses - future
Addresses Specify Network
Connections
• In general, an internet identifies a host
• If a router connects to two networks, then it
requires two IP addresses, one for each network
• Conventional computers that attach to more than
one network are called multi-homed hosts
Network Addresses
• A hostid of 0 (all zeroes) is not assigned to an
individual host, it indicates the network itself
• A hostid of all ones is a broadcast address and
refers to all hosts on the network (ones means all)
• When a packet is broadcast:
– a receiving router looks at the netid portion of the
address to pass it on to that network
– when the packet arrives at the destination network, that
router looks at the hostid; if all ones, it broadcasts
Limited Broadcast
• As part of a startup procedure, a host may use a
limited broadcast address (32 ones) before it
knows its IP address or the prefix for the local
network
• Generally, TCP/IP restricts broadcasting to the
smallest set of machines
Zeros and Ones
• A field consisting of ones can be interpreted to
mean “all”
• A field consisting of zeros can be interpreted to
mean “this”
– hostid = 0 is this host
– netid = 0 is this network
Addressing Extensions
• To conserve network prefixes, subnet addressing
allows multiple physical networks to share
prefixes
• In the 1990’s, it was determined that the division
between prefix and suffix could occur at any
point; this was called classless addressing, or
supernetting
• For right now, we just need to know that the
original scheme is no longer the most widely used
Weaknesses in Internet
Addressing
• Addresses refer to network connections, not the
computer itself
– If a host is moved from one network to another, its IP
@ must change
– Problem with mobile computers; when reconnected at a
remote site, it needs an IP @ at this new network?
• Class C networks can have no more than 255 hosts
• Which @ do we use if a host is multi-homed?
• See Figure 4.2
Dotted Decimal Notation
• IP addresses are written as four decimal digits,
separated by decimal points
• Each decimal corresponds to one octet in the 32bit IP address
• The 32-bit internet address
10000001 00001110 00001001 00011101
is written as 129.13.9.29
• See relationship of IP @ and dotted decimals in
Figure 4.3
Loopback Address
• 127.0.0.0 is used for testing software
• The protocol software on the computer interprets
this locally and sends no traffic across the network
• See summary of special addresses in Figure 4.4
Internet Addressing Authority
• ICANN (Internet Corporation for Assigned Names
and Numbers) sets policy and assigns values for
names and addresses
– Previously was IANA (Internet Assigned Number
Authority) until 1998 and was handled by Jon Postel
• This central authority assigns the network portion
of the address
• UHCL has 16 subnets 160 - 175. For example:
– 129.7.160.1
– 129.7. 173.15
Corporate Groups
• IBM has network address 9.0.0.0
• AT&T has 12.0.0.0
• If a corporation has no connections to the Internet,
it may use these addresses. Would it be wise?
• Why would someone have no connection to the
Internet?
• The IETF has reserved address prefixes and
recommends using them on private internets
Example from Purdue
• Mid 1980’s
• IP addresses are assigned to each network
connection
• Merlin has 2 addresses
• IP addresses for the routers (Glatisant and
Taliesyn) do not follow the convention
Network Byte Order
• Not all architectures store 32-bit integers the same
way
– Little Endian - low memory address holds low-order
byte of the integer
– Big Endian - low memory holds high-order byte
– How did you do it in Assembly?
• In TCP/IP, hosts and routers convert binary to
standard byte order before sending a packet, and
to host-specific order when receiving
• TCP/IP uses Big Endian
Summary
• TCP/IP uses 32-bit binary addresses (called IP
addresses) for machine identifiers
– address consists of a network id and a host id
• The classful scheme worked for a while, but had
to be extended
• IP addresses refer to network connections
• A standard byte-ordering scheme is necessary
For Next Time
• Read Chapter 5