Messages in this thread |  | | Date | Wed, 20 Nov 1996 11:01:23 GMT | From | Pedro Roque <> | Subject | Re: IPv6 and the "average user" |
| |
>>>>> "hpa" == H Peter Anvin <hpa@transmeta.com> writes:
hpa> Followup to: <3h7mnhjs36.fsf@Q.Net> By author: Bradley Ward hpa> Allen <ulmo@q.net> In newsgroup: linux.dev.kernel >> For instance, IPv4 has an address space of 2^32, or >> 4,294,967,296, addresses. That's a few billion less than there >> are people on the earth, or thereabouts, and we double every 60 >> years anyway (remember, half the people ever alive still are, >> or are dead, or however you prefer to look at two halves). At >> first glance 4,294,967,296 looks like a big number, but if your >> base is today's population, then it looks rather smallish (less >> than 1). >>
hpa> It is also important to point out that for various technical hpa> reasons, the actual number of possible hosts turn out to be hpa> much lower than that. Someone proposed a law saying that hpa> "the utilization of address space is inversely proportional hpa> to the address length"; although I personally think this law hpa> is overly optimistic, there is definitely a strong effect: as hpa> the address space goes larger, the more hierarchial it needs hpa> to become, and the less efficient the address utilization.
hpa> Either way, the results are the same -- 32 bits is just hpa> hideously insufficient. Personally I think it was a bad move hpa> of the IETF to pick a fixed-length standard for IPng (IP - hpa> The Next Generation), but my understanding is that router hpa> manufacturers (read: C*sc*) insisted...
Which is quite natural as nobody in it's right mind would want to route variable length addresses. Variable length fields should only exist in data to be processed in slow paths. Anyway, 128 bits is more than enought for IPv6. Note that the 32 bits where a good choice for IPv4.
regards, ./Pedro.
|  |