Class A Subnetting: Class A Subnet Experiment
A temporary experiment that carved class A network 39 into small pieces to flush out buggy routing software.
Internet Assigned Numbers Authority (IANA), "Class A Subnet Experiment", RFC 1797, DOI 10.17487/RFC1797, April 1995, <https://www.rfc-editor.org/info/rfc1797>.
@misc{rfc1797,
series = {Request for Comments},
number = 1797,
howpublished = {RFC 1797},
publisher = {RFC Editor},
doi = {10.17487/RFC1797},
url = {https://www.rfc-editor.org/info/rfc1797},
author = {Internet Assigned Numbers Authority (IANA)},
title = {{Class A Subnet Experiment}},
pagetotal = 4,
year = 1995,
month = apr,
} RFC text
Contents
Network Working Group IANA
Request for Comments: 1797 ISI
Category: Experimental April 1995
Class A Subnet Experiment
Status of this Memo
This document defines an Experimental Protocol for the Internet
community. This does not specify an Internet standard of any kind.
Discussion and suggestions for improvement are requested.
Distribution of this memo is unlimited.
Discussion
There appears to be some interest in experimenting with subnetting
the class A addresses.
There is some evidence that not all the routing software in use will
deal correctly with subnetted class A addresses. It also appears
that actual use of subnetted class A addresses may be necessary in
the not too distant future. It is suggested that conducting an
experiment now to identify and fix any software that does not
properly handle subnetted class A addresses would be useful and
important.
To further this experiment the IANA will temporarily designate the
class A network number 39 to be used in the following way:
The high order octet of the 4-octet IPv4 address is the class A
network number 39. There are two cases for low order 24 bits.
Case 1:
In the first case, the high order bit of these 24 bits is zero and
the next 15 bits are the low order 15 bits of a previously
assigned Autonomous System number (AS), as registered by a network
registry and listed in the RWhois database system.
Using the AS number in this way allows the experiment to get
underway quickly in that it automatically allocates some addresses
to each service provider and does not require a registration step.
One concern is that this might cause a run on AS numbers, since by
getting an AS number you automatically get some address space.
This concern should be offset by the fact that the amount of
address space one gets under this plan is the same as one class C
network number (and it should be easier to get a single class C
allocated than to get an AS number allocated), and that this is a
limited time experiment so that these addresses will be temporary.
The low order octet of the 4-octet IPv4 address is for local use.
It is expected that an address of this form will be used to
identify a specific publicly accessible Internet host.
+--------+-+-------+--------+--------+
| 39 |0| low 15 bits AS | local |
+--------+-+-------+--------+--------+
Case 2:
In the second case, the high order bit of these 24 bits is one,
and the remaining 23 bits are assigned by the IANA (currently
reserved for future use).
+--------+-+-------+--------+--------+
| 39 |1| variable prefix + local |
+--------+-+-------+--------+--------+
The general intent is to find a way to assign to experimenters
prefixes of differing lengths so that a variety of experiments can
be conducted with the prefix/local-address boundary at different
points.
It is not intended that either of these address allocation schemes is
the model for how subnetted class A addresses will be actually
allocated in the future.
It is expected, to make the experiment interesting, that some
providers will use these addresses for servers supplying popular
material via the Web or FTP.
For example, if the service provider registered to use AS 690 wished
to use this style of address to provide access to a server of popular
information on local host 7, the address would be:
+--------+--------+--------+--------+
| 39 | 2 | 178 | 7 |
+--------+--------+--------+--------+
The support for DNS name and address resolution should be provided.
For example, if Alternet wanted to put up a database of interesting
information using the hostname "Interesting.Alter.Net" using the
address 39.2.189.7, they would need to put the name to address
mapping in their name server using the A record
Interesting.Alter.Net. IN A 39.2.189.7
Similarly, the address to name PTR record should be supported
7.189.2.39.IN-ADDR.ARPA. PTR Interesting.Alter.Net.
which means that the 189.2.39 branch of the IN-ADDR tree would be
delegated to Alternet for the purposes of this experiment.
To support this, the 39.IN-ADDR.ARPA branch is delegated to the IANA
to be managed at ISI. The nameserver for this branch is
IN-ADDR.EP.NET (39.17.199.10). Participants in this experiment
should contact the administrator of this nameserver to have their
portion of the address space further delegated. The administrator
for this server can be reached at <aexpreg@isi.edu>.
Another aspect of the testing that should be performed is to have
providers interchange addresses to test the portability of subnetted
class A addresses. It is not intended that this would be the model
for actual use.
For example, if AS 690 and AS 1800 want to try out routing holes
in each others' allocations within their AS, that should be
encouraged. That is, suppose AS 690 handed some address of their
addresses to AS 1800, and vice-versa. This type of testing will
be necessary to see if the addresses can be made portable in
larger sub-A allocations.
This is experiment will be of limited duration and these addresses
may be reassigned to other uses when the experiment is over.
This experiment will begin on 1-May-95.
The current date for the termination of this experiment is 1-Dec-95.
Security Considerations
Security issues are not discussed in this memo.
Author's Address
Internet Assigned Numbers Authority (IANA)
Information Sciences Institute
University of Southern California
4676 Admiralty Way, Suite 1001
Marina del Rey, CA 90292-6695
Phone: 1-310-822-1511
EMail: iana@isi.edu
AI ELI5
On this page
§1Why Run the Experiment
There was interest in experimenting with subnetting class A addresses. The RFC says there is some evidence that not all routing software in use deals correctly with them, and that actual use of subnetted class A addresses might be necessary in the not too distant future.
So it suggests an experiment now, to identify and fix software that does not handle them properly. To do this, IANA temporarily designates the class A network number 39. The experiment begins on 1-May-95 and its current termination date is 1-Dec-95; the addresses may be reassigned to other uses afterwards.
Nobody knew whether some computers could cope with very big address blocks cut into small pieces, so they set up a trial to find out.
§2How Network 39 Is Divided
The high-order octet of the 4-octet IPv4 address is 39. The low-order 24 bits are used in one of two ways, chosen by their high-order bit.
Case 1: the bit is zero and the next 15 bits are the low-order 15 bits of a previously assigned Autonomous System (AS) number. The low-order octet is for local use, expected to identify a specific publicly accessible host. Tying the address to an AS number lets the experiment start quickly with no registration step. The space each AS gets equals one class C network, and the addresses are temporary.
Case 2: the bit is one and the remaining 23 bits are assigned by IANA (currently reserved for future use). The intent is to hand experimenters prefixes of differing lengths so the prefix and local-address boundary can sit at different points.
The RFC states that neither scheme is the model for how subnetted class A addresses will actually be allocated in the future.
The big block starts with the number 39. The rest of the number is built from a network's own ID plus a last piece the network can use for its own computers.
| Field | Name | Size | Value | Description |
|---|---|---|---|---|
NETWORK | Network | 1 byte | 39 | The class A network number. |
FLAG | Case bit | 1 bits | 0 | Zero selects Case 1. |
AS | AS number | 15 bits | Low-order 15 bits of an assigned AS number. | |
LOCAL | Local | 1 byte | Local use, such as a specific host. |
§3Examples and DNS Support
A provider registered as AS 690 that wants to serve popular information from local host 7 would use the address with octets 39, 2, 178 and 7. Providers are expected to put popular Web or FTP servers on such addresses to make the experiment interesting.
DNS name and address resolution should work too. For Interesting.Alter.Net at 39.2.189.7, the owner needs an A record, and a PTR record 7.189.2.39.IN-ADDR.ARPA.. That means the 189.2.39 branch of the IN-ADDR tree is delegated to Alternet. The 39.IN-ADDR.ARPA branch is delegated to IANA, managed at ISI, and participants contact that name server's administrator to get their portion delegated further.
Providers are also encouraged to swap addresses, for example AS 690 and AS 1800 trying routing holes in each other's allocations, to test whether subnetted class A addresses can be made portable. This is not meant as the model for actual use.
A company gets its own slice, names its computers inside it, and the phone book for names is set up to match. Companies can also trade slices to see if routers get confused.
§4Security and Status
The memo says only that security issues are not discussed in it. It is an Experimental protocol: it does not specify an Internet standard of any kind, and discussion and suggestions for improvement are requested.
The note does not talk about safety. It just says this is a trial, not a rule everybody has to follow.