RFC 917 Philippe Bissohong by 3J6W2pqN


									RFC 917                                                           Philippe Bissohong

                                   Internet Subnet

        Subnet is essentially the break down of one network. Organizations decide
for technical reason to split one network into multiple instead of getting many
network numbers. This RFC discuss the use of subnets, improvements and
problems in routing particularly.

        The original view of the internet was a two-level hierarchical: the catanet
(collection of connected internet networks) and below it, internet network numbers.
But it was later realized that it would be insufficient especially because big
organizations (Ex: universities or big companies) would need more than one LAN
cable to cover an area. So they added a third level, which divide one network into

When a company is faced with the problem of assigning internet address it has 3 choices.
The first one is acquiring distinct internet number for each cable but the disadvantage is
that it creates a really big routing table that is useless outside the organization. The
second thing it can do is use one internet network numbers but randomly assign host
number, the disadvantage for this one is that it will require that all technology support
ARP protocols or broadcasting but not all do. The third choice is subnet; the only
problem with subnet is, it changes the internet protocol, so the way IP is implemented
now must be change but it is a minor inconvenience with a good improvement to the

They propose in this rfc to use a standard internet address format over a variable or self-
encoding one. This is their proposal:
Where subnet number is constant for a given network, zero indicates no subnet. They
also propose that organization will choose subnet field width that is a multiple of 8 but
implement as well something that can handle any other width. They reject “recursive
subnet” or sub-subnet because of its complexity and the lack of bits in IP address.
Broadcasting should not only recognize its own host addresses but also three other kinds
of broadcast address:
Physical network: all ones ( will only send in the local physical network
Specific network: Contain actual network number while the rest is all ones.
Specific subnet: Contain actual network number and subnet number; all ones for host.

In order for a host to know the subnet field width in the network in which it is connected
or find a gateway on its local network or even find its own address, information needs to
be either “hardwired” or broadcast. Hardwired saves information in a disk file but it is
not likeable enough. As an alternative, since the majority of LAN technology supports
broadcasting, it is better for the host to just send a request for the information. For that
reason, they want to extend ICMP protocol, adding “Address format request” and
“Address format reply” to ICMP message types.

To top