On Thu, Nov 19, 1998 at 02:20:54PM +1300, Don Stokes wrote:
apnic-011 already says that, at least for addresses allocated in the NZGate timeframe. I haven't found any other APNIC document (expired or not) that states APNIC policy toward address ownership.
Assuming we all follow this then -- do we have allow people or users with small (eg. /26) networks to take them with them should they decide to move? Allowing this _without_ any constraints would make fragmentation horrific.
Do you have a clause in your service contracts that states explicitly what the position is regarding IP numbers you assign to clients? Most ISPs do (and all should).
No idea... I speak only for myself, not for any company. I don't look at the legal bits where possible, thats what marketroids and legal people are for.
Does it matter?
Yes. If someone is allocated address space for which they are not specifically told whether or not it should be considered portable or not, and therefore they wish to take the network with them when they move providers, we could get considerable fragmentation when many people with small networks do this.
It does if anyone is allocating address space in new blocks without explicitly stating the "ownership" of addresses, but for the old addresses it just means that at worst the routing table space taken up by old addresses doesn't get any smaller.
I don't follow; surely we have a situation where providers in the past have carved up a say /20 for clients -- and when a client moves this /20 might then need to become sixeteen /24 routes (or a /21, a /22, a /23 and two /24 or whatever).
It also matters if an ISP wants to move a bunch old /24 prefixes over to an upstream provider that refuses to deal with them -- so far the ones that have made noises about refusing small netblocks this have backed away from that position.
Eventfully as enough /24 are freed up, we will be able to coalesce adjacent ones into large networks.
What can be done as a technical group is to develop a consensus on how to retire old /24 prefixes and aggregate them into larger blocks, either through the APNIC's return policy (apnic-072), or through some local arrangement, without grossly impacting on either ISP or customer operations.
This seems like a good idea.
I think the first step in this is to get the "ownership" issues aired (I don't think it will be "solved") at the ISOCNZ conference tomorrow; at least then hopefully we'll have some idea of what various positions are. I don't think the ex-NZGate issue can be really proceeded on without that.
Can someone who attends this meeting then please provide some feedback? -cw --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Chris Wedgwood wrote:
On Thu, Nov 19, 1998 at 02:20:54PM +1300, Don Stokes wrote:
apnic-011 already says that, at least for addresses allocated in the NZGate timeframe. I haven't found any other APNIC document (expired or not) that states APNIC policy toward address ownership.
Assuming we all follow this then -- do we have allow people or users with small (eg. /26) networks to take them with them should they decide to move?
Allowing this _without_ any constraints would make fragmentation horrific.
Do you have a clause in your service contracts that states explicitly what the position is regarding IP numbers you assign to clients? Most ISPs do (and all should).
No idea... I speak only for myself, not for any company. I don't look at the legal bits where possible, thats what marketroids and legal people are for.
Does it matter?
Yes. If someone is allocated address space for which they are not specifically told whether or not it should be considered portable or not, and therefore they wish to take the network with them when they move providers, we could get considerable fragmentation when many people with small networks do this.
It does if anyone is allocating address space in new blocks without explicitly stating the "ownership" of addresses, but for the old addresses it just means that at worst the routing table space taken up by old addresses doesn't get any smaller.
I don't follow; surely we have a situation where providers in the past have carved up a say /20 for clients -- and when a client moves this /20 might then need to become sixeteen /24 routes (or a /21, a /22, a /23 and two /24 or whatever).
No. It's still a /20 with a single /24 hole punched in it. If the net ends up with a policy that only allows prefixes up to /20 on the backbone then the bozo who moves and won't renumber looses. He may 'own' the address space but it's no damn use to him. -- Mailto:Andy.Linton(a)netlink.net.nz Tel: +64 4 494 6162 Post: Netlink, PO Box 5358, Lambton Quay, Wellington, New Zealand -- --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Thu, Nov 19, 1998 at 08:37:28PM +0000, Andy Linton wrote:
No. It's still a /20 with a single /24 hole punched in it. If the net ends up with a policy that only allows prefixes up to /20 on the backbone then the bozo who moves and won't renumber looses. He may 'own' the address space but it's no damn use to him.
You missed my point, a /20 with a /24 hole in it means we end up having to make multiple routes where there was one. This becomes more work for administrators when creating access lists, etc. Arguably, you can leave the /20 and create a /24 with a different gateway in some circumstances, but this I beleive is also going to be error prone. -Chris --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Chris Wedgwood wrote:
On Thu, Nov 19, 1998 at 08:37:28PM +0000, Andy Linton wrote:
No. It's still a /20 with a single /24 hole punched in it. If the net ends up with a policy that only allows prefixes up to /20 on the backbone then the bozo who moves and won't renumber looses. He may 'own' the address space but it's no damn use to him.
You missed my point, a /20 with a /24 hole in it means we end up having to make multiple routes where there was one. This becomes more work for administrators when creating access lists, etc.
We end up with one additional route.
Arguably, you can leave the /20 and create a /24 with a different gateway in some circumstances, but this I beleive is also going to be error prone.
Arguably, I'd argue that's precisely what you do. -- Mailto:Andy.Linton(a)netlink.net.nz Tel: +64 4 494 6162 Post: Netlink, PO Box 5358, Lambton Quay, Wellington, New Zealand -- --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Mon, Nov 23, 1998 at 05:22:09AM +0000, Andy Linton wrote:
We end up with one additional route.
This assumes all routers know about all routes at all times -- which isn't necessarily the case. It also make things like access-lists more complex. -cw --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Fri, 27 Nov 1998, Chris Wedgwood wrote:
On Mon, Nov 23, 1998 at 05:22:09AM +0000, Andy Linton wrote:
We end up with one additional route.
This assumes all routers know about all routes at all times -- which isn't necessarily the case.
No, it assumes that apart from the core routers, each router has a default route and a list of exceptions to that default.
It also make things like access-lists more complex.
Agreed. Which is why we're meeting today to try to sort out the portablity issues. --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Thu, Nov 26, 1998 at 08:25:31PM +0000, Andy Linton wrote:
No, it assumes that apart from the core routers, each router has a default route and a list of exceptions to that default.
You peoplel must be smarter than me -- I honestly can say from experience I've had trouble with situations like this is the past.
Agreed. Which is why we're meeting today to try to sort out the portablity issues.
See ya soon then.... -cw --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Thu, 19 Nov 1998, Andy Linton wrote:
Chris Wedgwood wrote:
I don't follow; surely we have a situation where providers in the past have carved up a say /20 for clients -- and when a client moves this /20 might then need to become sixeteen /24 routes (or a /21, a /22, a /23 and two /24 or whatever).
No. It's still a /20 with a single /24 hole punched in it. If the net ends up with a policy that only allows prefixes up to /20 on the backbone then the bozo who moves and won't renumber looses. He may 'own' the address space but it's no damn use to him.
I may be missing the point here, but in a situation where the net won't route longer prefixes than /20, doesn't punching a /24 hole in a /20 render the entire /20 unusable, not just the /24 - because the carrier who formally advertised the /20 now has to advertise a bunch of /21,/22,/23 and /24 prefixes to cover the remainder? Cheers Si --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Fri, Nov 20, 1998 at 10:02:21AM +1300, Simon Blake wrote:
I may be missing the point here, but in a situation where the net won't route longer prefixes than /20, doesn't punching a /24 hole in a /20 render the entire /20 unusable, not just the /24 - because the carrier who formally advertised the /20 now has to advertise a bunch of /21,/22,/23 and /24 prefixes to cover the remainder?
Yes... hence why where possible I believe people with small networks should be encouraged to renumber when the moving providers. Right now, /24s are accepted almost everywhere and it doesn't look like anyone is going to refuse them anytime soon though but I still don't think that means we can be complacent. -cw --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Simon writes:
I may be missing the point here, but in a situation where the net won't route longer prefixes than /20, doesn't punching a /24 hole in a /20 render the entire /20 unusable, not just the /24 - because the carrier who formally advertised the /20 now has to advertise a bunch of /21,/22,/23 and /24 prefixes to cover the remainder?
No. The /20 supernet continues to be advertised, and the new /24 net is also advertised. Since the /24 advertisement is more specific, traffic is routed to it; if it's not inside the /24 but inside the /20, the less specific route is used. For example, if carrier A advertises 192.168.0.0/20 and carrier B advertises 192.168.4.0/24, carrier C would have two routing table entries: 192.168.0.0/20 -> a.a.a.a 192.168.4.0/24 -> b.b.b.b and traffic for 192.168.4.2 would go to the more specific route b.b.b.b, while traffic for 192.168.5.4 wouldn't match the /24, and instead takes the less specific /20 route to a.a.a.a. It doesn't cause the /20 to be split into individual routes, but does create one extra routing table entry for each "hole". -- don --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
participants (5)
-
Andy Linton -
Chris Wedgwood -
Chris Wedgwood -
Don Stokes -
Simon Blake