Re: Raising and lowering the roof of the DNS;-)...

["Jim Fleming" <[e-mail]>]

----- Original Message ----- 
From: "Einar Stefferud" <[e-mail]>
> 
> Now for the other issue:  The argument about whether the Brits had it
> right with their big Endian logic, vice the US little Endian logic,
> mostly boils down to whether we are all driving on the same side of
> the road, or not.
> 

Stef,

This is not rocket science. There are 160 bits in most IPv4 headers.
That is 5 (fiive) 32-bit words. It is a data structure. Any 8th grade
computer programmer can define the fields and lay out the bits. Any
Internet technologist should be able to tell you what all 160 bits are
in less than 1 minute. You might want to walk up to a person at an
IETF meeting, and ask them to run thru the fields. See if they can even
get them in the right order. Would you expect to ask a chemist about
the periodic table of the elements, and find they do not know it, cold ?

There are 128-bits in the 128-bit DNS system. Hmmm, wonder how
that happened ? The 32-bit DNS system has 32-bits. 32 is smaller than
128, at least for most people. You can go to a remote region and place
32 stones in one pile and 128 stones in another pile, and most people
would agree the pile with 128 is larger. You can build on these basic
UnirVersal truths. If you find someone who disputes it, you are likely
wasting your time.

Imagine a simple DNS to IPv4 header mapping. Take the 128-bits from
the 128-bit DNS and toss them into the first 128 bits of the 160 bit IPv4
Header, and send the packet. One problem, most humans agree, that
128 is less than 160 bits, so you would not fill all the places. Does that
mean you should advocate for a 160-bit DNS ? You could try. You might
also sit back and figure out how many of the 160 bits can be "touched" or
set via the DNS. How fast can you answer that question ? You might want
to walk up to a person at an IETF meeting and ask them that question.
Can they reason through the fact that the 4-bit Version field is set at 0100,
the 4-bit header length is set at 0101 (5), the 8-bit TOS field is usually 0,
the 16-bit Length field can not be touched...etc. etc. etc. In other words,
in the first 32-bits, can they instantly note that only 8 of the bits can be
touched ? 160-32=128...can you figure out which of the other 128 bits
can be touched ?

Are you starting to see that with 128-bits, and a 160-bit header, 128-bits
allow people to touch all of the fields they can touch with bits left over.
Now what does one do ?...the 128 bits do not fit in the 160 bits...because
not all 160 bits can be set via the 128-bit DNS and have the EXISTING
routers and servers all work...

Looking at a bigger picture starting point, one sees that UDP and TCP
dominate scene beyond the first 160 bits....and there are 16 bits of port
number that would be nice to be able to set....that was not possible with
32-bit DNS...all 32-bits got blindly tossed into fields of the 160 bits in
the trivial, "toy", 32-bit Internet....

If you look at the 128-bits as a starting point for a new computer architecture
addressing scheme, you will quickly see that you will likely want to save some
room for the 16-bit UDP/TCP Port number....you are then down to 112 bits...
You likely will also want to give people a 32-bit IPv4-like address field to
help them preserve their look and feel and intranet addressing...112-32=80
You now have 80 bits and might want to consider taking 64 of those for
various transition mechanisms, and opcodes, etc. That leaves you with 16 bits
because 80-64=16....you now have 16 bits to try to put into the EXISTING
IPv4 header....can you do it ?....and still have all routers continue to work...?

Does the IETF do technical work ?

Jim Fleming
2002:[IPv4]:000X:03DB
http://www.iana.org/assignments/ipv4-address-space
http://www.ntia.doc.gov/ntiahome/domainname/130dftmail/unir.txt