10 * 172.16 * 192.168 ... why?
in [Linux Networking]
|
From: antonio on 13 Apr 2008 13:10 Moe Trin ha scritto: > On Sun, 13 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in > article <4802096d$0$29603$4fafbaef(a)reader1.news.tin.it>, antonio wrote: > >> does have any sense this numbers? > > It was what was available. > >> 10, may be, it's easy to remember for our decimal-sistem > > /8 (formerly called Class A) address ranges are limited. There were > only 128 in the original scheme, and they threw away two of those > (0.0.0.0/8 and 127.0.0.0/8) leaving 126. Today, 96 of them have been > assigned. See http://www.iana.org/assignments/ipv4-address-space. In > fact, 10.0.0.0/8 was the old DARPA address range and it was retired. > See RFC1627 for indications of the problems about the re-use of that > address range. > >> 192 it's the first of the C class... but why *168*? >> and 172.16? >> >> it seems random... > > It's what was available. See RFC3330 for further hints. Address space > is in demand, which is why IPv6 was developed. As of last month, there > were 2609197794 IPv4 addresses in use - 70.4 percent of the available. > For perspective, the same figures for last December (3 months earlier), > were 2564407724 and 69.2 percent. At the end of 1993, about the time of > RFC1597 (which was replaced by RFC1918), the figures were 1233944064 and > 33.3 percent, so one third of the address space was already in use. > > Old guy WOW! amazing!
From: antonio on 13 Apr 2008 14:26 > It's what was available. See RFC3330 for further hints. Address space > is in demand... ....and maybe the most *sorted* adresses was demanded first...
From: ray on 13 Apr 2008 20:41 On Sun, 13 Apr 2008 15:18:46 +0200, antonio wrote: > hi > does have any sense this numbers? > > 10, may be, it's easy to remember for our decimal-sistem 192 it's the > first of the C class... but why *168*? and 172.16? > > it seems random... By my calculations, 10 * 172.16 * 192.168 = 330836.42880.
From: Moe Trin on 14 Apr 2008 16:11 On Sun, 13 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in article <48025192$0$29596$4fafbaef(a)reader1.news.tin.it>, antonio wrote: >> It's what was available. See RFC3330 for further hints. Address space >> is in demand... > >...and maybe the most *sorted* adresses was demanded first... That is more difficult to say. A quick grep of the delegation files from the five RIRs looking for IPv4 addresses in the 192.x.x.x range that were allocated/assigned before the end of 1993 turns up 12166 lines. These allocations/assignments are coming from three RIRs (APNIC, ARIN, and RIPE because AFRINIC and LACNIC didn't exist until later), and it appears that they were allocating space individually. Each person doing these allocations/assignments was following their own procedure within those blocks - such that one may have started at the beginning of his block and assigned consecutive ranges, while another may have been trying to assign blocks based on the region or country. (Recall that originally, ARIN was making all allocations/assignments, then RIPENCC took over for Europe and Africa, then APNIC took over the Asia/Pacific part of the world, eventually followed by LACNIC for Central/South America and most recently, AFRINIC for Africa.) This more or less is the same "problem" about how do you assign IP addresses on your LAN. Do you start at x.y.z.1 and work up? Or have you some numbering scheme, where hosts x.y.z.1 through x.y.z.9 are routers and DNS servers, hosts x.y.z.10 through x.y.z.N are workstations located on the first floor or this building or some similar idea? Old guy
From: Moe Trin on 23 Apr 2008 16:02 On Tue, 22 Apr 2008, in the Usenet newsgroup comp.os.linux.networking, in article <gk32e5-4ca.ln1(a)royalty.mine.nu>, Hactar wrote: >Moe Trin <ibuprofin(a)painkiller.example.tld> wrote: >>Hactar wrote: [RFC1918 addresses] >>>They require relatively few components to recognize in hardware: >> >> Ummm, when was the last time you popped the hood on a computer, and >> had a look at the components on the motherboard? Looking for IP >> addresses out of the billions of items of data that are floating >> around doesn't require dedicated hardware. It's done as a bit >> comparison in the CPU. > >_Now_ it is. How were routers constructed when IPv4 was designed? That would be back in 1980, and the few routers in existence at the time were based on mini-computers - try a PDP-11 as an example. Simple reason. Assume a router with three network interfaces. Which interface are you going to send the packets destined for network host 10.10.10.10 to? Is it going to be the same one as 10.20.30.40? Obviously each installation is going to have there own choice of addresses, located on the interface of their choice. Are you going to have a hardware comparator for each network range? How are you going to configure them - DIP switches? (It's not easy to decode the locations, but have a look at RFC0846, and see where the _hosts_ with IPs 10.1.0.x were located in early 1983. You might recognize MIT and USC as being at opposite sides of the continent.) Not that many people were building their own routers, any more than people were building their own network cards. Consequently, those routers had to be programmable to cater for the different IP ranges in use. If you stop and think of the hardware required (those packets could be 1526 octets long which includes the preamble of 8 bytes, Ethernet header of 14 bytes [and trailer of 4 bytes]) and the destination IP address is going to be 4 bytes starting 16 or 24 bytes past that 14 byte header depending on the version of Ethernet encapsulation (RFC0894 verses RFC1042, verses several proprietary versions), implementing this in hardware is going to be "interesting" - that's sure a bunch of 74LS373s never mind the circuitry that is going to clock each one. And don't forget that your router is decrementing the TTL (13th byte in the IP header) and you have to recompute the header checksum (15th and 16th byte in the IP header), and while you're at it, you need to look at the IP options (if any) starting at the 21st byte, to see if there isn't something like routing commands or a 'record-route' option present (meaning you've more more work to do with the packet before shipping it to the next hop). The last problem is that IP wasn't the only network protocol that was running on those wires - even back in 1980. And worse, a few of those protocols were routable. Old guy
First
|
Prev
|
Pages: 1 2 Prev: Share clipboard and map network drive on rDesktop Next: Ping and Tracert |