• Welcome to Hurricane Electric's IPv6 Tunnel Broker Forums.

What do you call the 16-bit groups of an IPv6 address?

Started by dstickneywork, August 04, 2009, 02:50:06 PM

Previous topic - Next topic

dstickneywork

Hey all,

I've been working with IPv6 for a while now, and I forget my original source, but I've been calling the eight colon separated 16-bit groups of an IPv6 address "words". I see some other sources that use the same terminology (one example is http://documents.iss.net/whitepapers/IPv6.pdf), but I see others calling them "octets". Octets doesn't seem accurate to me because an octet is an 8-bit group ( http://en.wikipedia.org/wiki/Octet_(computing) ), like in IPv4 addresses.

I am interested to find out what other enterprise IT/network engineers call them. Does anyone know of an official source (RFC, IETF, ARIN, etc) that defines what they are called?

Thanks,
-Daniel

broquea

Hexadecitet or double-octet or something involving nybbles?

dstickneywork

Really? Do you use any of those words in a professional context when talking with another network engineer? It seems that for such an important part of the IPv6 vocabulary there ought to be a common and official term. I thought the term "word" was official. I'd love to hear more input on the question.

Thanks!

broquea

Quote from: dstickneywork on August 04, 2009, 06:13:04 PM
Really? Do you use any of those words in a professional context when talking with another network engineer? It seems that for such an important part of the IPv6 vocabulary there ought to be a common and official term. I thought the term "word" was official. I'd love to hear more input on the question.

Thanks!

Oh, I was just tossing ideas out there. I've never referred to the bits as that in a professional manner, only joking around with others as to what to call them. I'd say go with whatever the IPv6 RFCs call them.

jimb

I call them words too, although it's somewhat imprecise since "word" don't necessarily imply length, although to myself and other programmers I've worked with in the past, a short word (short), or just "word" connoted 16 bits, and a long word (or long) is 32, etc.  Gotta love ambiguous terms, eh?

Octets is clearly wrong, although double-octet or double-byte would be right, but a bit of a mouthful and a pain to type.  Nybble is clearly wrong, since it's four bits (or half of a byte ... get it?).

Maybe they should use a generic term like "elements", although that smacks of arrays. 

The stuff inside your colons?   :D (yeah I know...baaaad joke)

snarked

They are referred to as "quads" - because in their text form, they have 4 hex digits.

Word size is a VARIABLE size that depends on the native number of bits of the CPU a machine uses.  A word is 64-bits on a 64-bit machine, 32-bits on a 32-bit machine, etc....

jimb

Yeh as I said "word" is a bit imprecise, but still works.  Back in the day (showing my age here), when we had 8 and 8/16 processors, word often implied 16 bits by default unless prefixed by something else or specified, at least among my crowd.  But yeah, as I said, technically it doesn't imply a particular word size.  But that doesn't make it "wrong" to use for IPv6 address components.  

Quad is fine with me too, although might be confused with the old "dotted quad" by users, although the "dotted" part should provide a clue.  :P

Have you seen them called quads in a lot of docs and such?

dstickneywork

Quote from: broquea on August 04, 2009, 06:21:29 PM
Oh, I was just tossing ideas out there. I've never referred to the bits as that in a professional manner, only joking around with others as to what to call them. I'd say go with whatever the IPv6 RFCs call them.

Ahh, ok :)  Thanks for the input (and the HE tunnel service!). I am preparing the execution of our enterprise wide IPv6 integration project plan and discussing IPv6 with other senior network engineers, and I want to make sure I'm using the official and correct term, if there is one. I've skimmed through the address format specifications in the RFCs and was unable to find a single term for them. Most places that describe the address structure call them some huge name like "eight groups of hexadecimal quartets separated by colons". RFC 1884 calls them "the eight 16-bit pieces of the address" (http://tools.ietf.org/html/rfc1884#page-4).

One of the more authoritative places I've seen the term "word" used is in the book The TCP/IP Guide (ISBN-13: 9781593270476) page 379. The IBM Internet Security Services link in my first post also holds a good amount of weight.

Thanks Jimb and snarked for your comments! Snarked: can you provide some authoritative references which demonstrate them being called "quads"? I like that it is a single short word, but I need to make my decision based exclusively on authoritative sources. Also I agree that "word" is variable, and potentially not the best term. Anyone have friends/contacts in the IETF who can be a reference?

jimb

LOL nomenclature.  Quartets eh?  That works too.  Maybe abbreviate that to quarts.  But then that sounds like you're talking about containers of milk or something.

Also, it sort of grates against "octets" which derives from the fact that a byte has eight bits, where "quartets" seems to derive from having four nybbles or characters/hex digits.  

The seemingly "proper form" following the derivation of "octet" would seem to be "hexadecatet", according to http://en.wikipedia.org/wiki/Numerical_prefix.   What a mouthful!  I'd be good with abbreviating it "hextets" though.  :P

No matter what you pick, there will probably be some de facto standard that comes into common use that you will be forced to use no matter how good/bad it is.  Kinda like how consumer level users call just about everything network related a router.  

Also, I found an interesting article (for geeks) that shows the timeline of some of this nomenclature:  http://en.wikipedia.org/wiki/Timeline_of_binary_prefixes

It shows how some manuals back in the 70s/80s used "words" to imply 16 bits as I was talking about earlier, even though it was probably improper use.

:D

broquea

Ah HA! I knew I saw hexadeca- somewhere before. I've been staring blankly at http://en.wikipedia.org/wiki/Hexadecimal#Cultural as well figuring out something with greek/latin in it.

jimb

Quote from: broquea on August 05, 2009, 05:40:00 PM
Ah HA! I knew I saw hexadeca- somewhere before. I've been staring blankly at http://en.wikipedia.org/wiki/Hexadecimal#Cultural as well figuring out something with greek/latin in it.
Haha.  Interesting article.  According to that, proper terms might possibly include "sexadecitet", or "senidentet" (Knuth).   :D

Sometimes I wonder if it was wise to settle on the colons and 16-bit hex words to represent IPv6 addresses.  It's great for IT/CS types since hex is so easy to convert to binary and makes identifying subnets easier, etc (since hex digits always fall in even four bit boundaries, unlike decimal digits).  But it might be daunting for non technical people.  They could have kept the "dotted quad" notation and increased the range from 0-255 to 0-4294967295 (ex: 12343243.1342.43213.27), but that would likely cause confusion with IPv4 addresses.  Or they could have decided on eight groups of 0-65535 (ex: 43234.1234.6674.0.7831.8343.325.99), but that's kinda long, and not much diff than eight groups hex digits.  Meh.  I guess it's the best compromise.   :-\

snarked

Some call "AAAA" as "quad-A", so it carries over.   (Source:  http://www.isi.edu/~bmanning/v6DNS.html )

Quad seems to be any grouping of 4.

A precise reply:

  "1. IPv6 addresses are 128 bits, written as eight quads of hex digits. Each quad is delimited by a colon."

(Source: http://www.gossamer-threads.com/lists/nsp/ipv6/18174 post #8)

However, absent from http://tools.ietf.org/search/draft-kawamura-ipv6-text-representation-02 (draft) -> RFC 4291 -> ... -> RFC 1884.

jimb


devroot

On a related note. Has anyone come across a nomenclature for the double colon :: Seems there should be a better way to say it than just "colon colon" or "double colon" 

dataless