Re: About address allocating (IPv6, variable length SLA/prefixes?)
Architecturally, IPv6 has the equivalent of variable length subnet masks built in. There are really only two boundaries that are not flexible- the boundary between the format prefix and the rest of the address, and the /64 boundary. (The format prefix is in fact variable length, but it is architecturally defined.) So any IGP or EGP design needs to be fully flexible to the left of /64. Subnetting to the right of /64 would be tricky.
all IGP's and EGP's have to support all prefix-lengths. e.g you want to announce host routes, /96 for NAT-PT, etc. for that matter, the implementation(s) I know will let you create subnets of whatever size you like. /ot
Ole Troan wrote:
Architecturally, IPv6 has the equivalent of variable length subnet masks built in. There are really only two boundaries that are not flexible- the boundary between the format prefix and the rest of the address, and the /64 boundary. (The format prefix is in fact variable length, but it is architecturally defined.) So any IGP or EGP design needs to be fully flexible to the left of /64. Subnetting to the right of /64 would be tricky.
all IGP's and EGP's have to support all prefix-lengths. e.g you want to announce host routes, /96 for NAT-PT, etc. for that matter, the implementation(s) I know will let you create subnets of whatever size you like.
Yes, but there are other problems if you subnet to the right of /64 (autoconfiguration for example). I said tricky, not impossible. Brian
At 10:06 AM -0500 5/22/00, Brian E Carpenter wrote:
Yes, but there are other problems if you subnet to the right of /64 (autoconfiguration for example).
It depends on the address format. Not all format prefixes require the 64-bit interface ID field. Agreed that, in the global aggregatable format, subnet prefixes longer than /64 create problems. However, implementations should be kept unaware of the magic /64 boundary, as much as possible. (Obviously, the piece of code that does stateless autoconfig has to know about it, but certainly not any routing code.) Steve
Steve Deering wrote:
At 10:06 AM -0500 5/22/00, Brian E Carpenter wrote:
Yes, but there are other problems if you subnet to the right of /64 (autoconfiguration for example).
It depends on the address format. Not all format prefixes require the 64-bit interface ID field.
Agreed that, in the global aggregatable format, subnet prefixes longer than /64 create problems.
However, implementations should be kept unaware of the magic /64 boundary, as much as possible. (Obviously, the piece of code that does stateless autoconfig has to know about it, but certainly not any routing code.)
I didn't mean to imply otherwise. Brian
participants (3)
-
Brian E Carpenter -
Ole Troan -
Steve Deering