Messages in this thread |  | | Date | Tue, 3 Dec 1996 20:37:31 -0800 (PST) | From | Jason Eggleston <> | Subject | IPv6 and the Future |
| |
I've been silently reading the IPv6 thread, and decided to speak up. I've listed my limited observations on IPv6, to help explain it to some people on this list (mostly the ones who were lurking like me, and who probably haven't read much on it.) My question, which is not a nuts and bolts kernel question, is at the bottom.
I just read a book on IPv6 (won't mention the title). The general impressions I observed were:
1. Increased address size for: a. massive address supernetting (exponentially smaller routing tables) based on what amounts to ISP prefixes. The idea is to have one prefix per ISP, which would translate to 1, or less than 1, roting entry per ISP in backbone routers. b. each way to access a host will have a seporate IP address. For example, if a host is connected through multiple ISPs... and an address that is only vaild on the LAN. 2. Various updates to ICMP which support multicast routing, peer discovery (without ARP), and router discovery. 3. My main question here: flow lables. I've been working a lot with frame relay and have been learning some things about ATM. I was pretty much sold on the idea of ATM. But, after reading the book, I learned just how bad the war between switching and routing is.
I work at an ISP that happens to favor switching. I've heard serious reasons why switching would be better than routing. (and several definitions of 'better'). I'm looking for a good argument in favor of routing. The 'stateless' arguments I've seen don't seem like much of a plus if every packet has to be routed.
I apologize if this message is not fit for this list. It does seem more informative than some of the other ones I've seen, and fairly relevant to one of the ongoing threads here. Also, IPv6 is still in development for the kernel.
Thanks,
Jason Eggleston
|  |