How did the private ip address ranges get selected?
January 16, 2009 9:32 AM Subscribe
How did 192.168 and 10, etc get selected as the private IP address ranges? I know it's in the RFC, but is there a story about why they picked those particular numbers? Who picked them?
Response by poster: I'm pretty sure that's not the case.
posted by empath at 9:48 AM on January 16, 2009 [1 favorite]
posted by empath at 9:48 AM on January 16, 2009 [1 favorite]
I'm going to make a guess - for 10/8, IANA wanted to create a reserved address space, so they just looked at all the /8's available, and said "10's available, let's use that"
posted by TravellingDen at 9:56 AM on January 16, 2009
posted by TravellingDen at 9:56 AM on January 16, 2009
... and why does 172.16.0.0/12 never get the love that it deserves?
posted by togdon at 10:18 AM on January 16, 2009 [2 favorites]
posted by togdon at 10:18 AM on January 16, 2009 [2 favorites]
Best answer: 10.0.0.0/8 was the old ARPANET, which they picked up on 01-Jan-1983. When they shut down the ARPANET in 1990, the 10.0.0.0/8 block was freed. There was much argument about if there should ever be private IP spaces, given that a goal of IPv4 was universal to all hosts on the net.
In then end, practicality won out, and RFC 1597 reserved the now well known private address spaces. When ARPANET went away, the 10.0.0.0/8 allocation was marked as reserved and since it was known that the ARPANET was truly gone (the hosts being moved to MILNET, NSFNET or the Internet) it was decided that this was the best Class A block to allocate.
Note Class A. This was before CIDR. So, the Class A, B and C private address netblocks needed to come out of the correct IP ranges.
I know that 172.16.0.0/12 was picked because it offered the most continuous block of Class B (/16) addresses in the IP space that was in a reserved block. 192.0.0.0/24 was always reserved for the same reason that 0.0.0.0/8 and 128.0.0.0/16 were reserved (first blocks of the old Class C, A and B network blocks) so assigning 192.168.0.0/24 out as private fit well -- 192.0.2.0/24 was already TEST-NET, where you could use them in public documentation without fear of someone trying it (see example.com for another example.)
posted by eriko at 10:51 AM on January 16, 2009 [28 favorites]
In then end, practicality won out, and RFC 1597 reserved the now well known private address spaces. When ARPANET went away, the 10.0.0.0/8 allocation was marked as reserved and since it was known that the ARPANET was truly gone (the hosts being moved to MILNET, NSFNET or the Internet) it was decided that this was the best Class A block to allocate.
Note Class A. This was before CIDR. So, the Class A, B and C private address netblocks needed to come out of the correct IP ranges.
I know that 172.16.0.0/12 was picked because it offered the most continuous block of Class B (/16) addresses in the IP space that was in a reserved block. 192.0.0.0/24 was always reserved for the same reason that 0.0.0.0/8 and 128.0.0.0/16 were reserved (first blocks of the old Class C, A and B network blocks) so assigning 192.168.0.0/24 out as private fit well -- 192.0.2.0/24 was already TEST-NET, where you could use them in public documentation without fear of someone trying it (see example.com for another example.)
posted by eriko at 10:51 AM on January 16, 2009 [28 favorites]
Response by poster: Eriko, that's such a good answer, you should put it in wikipedia. That was the first place I looked, and it seems like it should be there.
posted by empath at 10:56 AM on January 16, 2009
posted by empath at 10:56 AM on January 16, 2009
Seconding good answer / wikipedia. I spent a good 1/2 hr poring through the RFC, other docs trying to find something substantial.
eriko --while this all makes sense do you have a source or are these details you've gleaned over the years?
posted by ezekieldas at 11:05 AM on January 16, 2009
eriko --while this all makes sense do you have a source or are these details you've gleaned over the years?
posted by ezekieldas at 11:05 AM on January 16, 2009
eriko, awesome answer... I'm curious where you found that, or if you were just around to live it, as I'm sure I wasn't the only one to have (re-)read RFC 1918 in its entirety today.
posted by togdon at 3:36 PM on January 16, 2009
posted by togdon at 3:36 PM on January 16, 2009
I'm a bit of an old phart. :-)
This wasn't my time, but it was just before my time -- I remember the *big deal* that CIDR was. Before that, Internet routers didn't need netmasks -- local routers might, because you might be subnetting, but a public address had it's netmask implicitly assigned by what class it was in. CIDR is what we always wanted, but the memory requirements would have slaughtered the routers of the time.**
The Private IP spaces were a big argument -- note that RFC 1918 explicitly obsoletes two others, RFC 1627 and 1597. Read those for a sense of the friction in the days when the net was moving from research toy to The Internet.
Indeed, a great deal of the history of the net can be gleaned from the RFCs. RFC, after all, does stand for "Request For Comments." There's a surprising amount of history there. Take a guess as to what the date of RFC 1 is.
** Last I checked, April 2008, to have the full routing table would required you to support 250,000 routes, in 39,000 Autonomous Systems, and that's just the public side. A big ISP will half half again that in private routes. Thankfully, memory is basically free now.
posted by eriko at 7:42 PM on January 16, 2009
This wasn't my time, but it was just before my time -- I remember the *big deal* that CIDR was. Before that, Internet routers didn't need netmasks -- local routers might, because you might be subnetting, but a public address had it's netmask implicitly assigned by what class it was in. CIDR is what we always wanted, but the memory requirements would have slaughtered the routers of the time.**
The Private IP spaces were a big argument -- note that RFC 1918 explicitly obsoletes two others, RFC 1627 and 1597. Read those for a sense of the friction in the days when the net was moving from research toy to The Internet.
Indeed, a great deal of the history of the net can be gleaned from the RFCs. RFC, after all, does stand for "Request For Comments." There's a surprising amount of history there. Take a guess as to what the date of RFC 1 is.
** Last I checked, April 2008, to have the full routing table would required you to support 250,000 routes, in 39,000 Autonomous Systems, and that's just the public side. A big ISP will half half again that in private routes. Thankfully, memory is basically free now.
posted by eriko at 7:42 PM on January 16, 2009
« Older What percentage of U.S. power plants can run on... | What's the best software for cataloging an art... Newer »
This thread is closed to new comments.
posted by zippy at 9:40 AM on January 16, 2009