Using /126 for IPv6 router links

Mark Smith nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org
Wed Jan 27 14:41:30 UTC 2010


On Thu, 28 Jan 2010 00:26:34 +1100
Mark Andrews <marka at isc.org> wrote:

> 
> In message <m2sk9rsobb.wl%randy at psg.com>, Randy Bush writes:
> > >>> the general intent of a class B allocation is that it is large enough
> > >>> for nearly everybody, with nearly everybody including all but the
> > >>> largest of organisations.
> > >> That would, indeed, work if we weren't short of class B networks
> > >> to assign.
> > > Would you clarify? Seriously?
> > 
> > we used to think we were not short of class B networks
> 
> Really?  Do you have a citation?  It should have been clear to
> anyone that thought about it that IPv4 address where not big enough
> to support every man and his dog having a network.
> 

If you dig into it a bit, you find that the original addressing plan
was a single network octet, and 3 node octets. The earliest document I
can find that describes 32 bit IP addresses is Internet Engineering Note
5, March 1977 (http://www.isi.edu/in-notes/ien/ien5.pdf), page 68 (69 of
the .pdf), and a diagram on page 74 (page 75 .pdf).

IEN 91, May 1979 (http://www.isi.edu/in-notes/ien/ien91.txt), also
describes the earliest 32 bit IP address format, and how to map link
layer addresses, such as ARPANET addresses into the "Local address"
portion.

RFC760, January 1980, also specifies that format of addressing. RFC791,
September 1981, is where it changed to classes. So IP addresses were
structured and deployed with a single network octet and 3 node octets
for more than 4 years.

I think that is evidence that 32 bit IP addresses were never originally
designed to support a world wide network that the Internet has become,
that fixed network and node portions are the preferred way to do
network addressing (and if you look at all the other protocols that
have existed, excepting CLNS (a "fixed" copy of IPv4 apparently),
they've all done it that way), and that classes, subnets and then
classless addressing have all fundamentally been very been neat hacks
to make 32 bit addressing support far more devices than was ever
expected.

> I know when I was getting my first class B address block in '88
> that it was obviously not sustainable but I'll get one while I can
> because that and class C's were all that were available and it could
> be justified under the rules as they stood then.
> 
> CIDR when it came along didn't change my opinion, though it did
> delay the inevitable as did PNAT.
> 
> I don't see the same thing with /48 as the basic allocation provided
> RIR's don't do greenfield all the time but instead re-allocate
> blocks when they are not maintained.  Always doing greenfield
> allocations will exhaust any allocation scheme in time.
> 
> Mark
> 
> -- 
> Mark Andrews, ISC
> 1 Seymour St., Dundas Valley, NSW 2117, Australia
> PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org
> 




More information about the NANOG mailing list