ARIN about to make the routing table a whole lot fatter?

[Francisco]

So there's a proposal on ARIN's desk to allow the transferring/selling of < /24 allocations.

Are we about to see a multi million entry routing table in the coming couple years?

People with older switches that top at 512k routes are having a hard enough time as it is,

but even > 1M route routers may not be enough.

I'm hoping ARIN turns it down in the end since it's going to turn really ugly, or just a lot

default routes are going to start popping up.

You can read up on the policy over at https://www.arin.net/policy/proposals/2014_3.html

There's a discussion happening over at http://lists.arin.net/pipermail/arin-ppml/2014-April/028245.html

Francisco

 

[SkylarM]

Not the best of ideas. No thanks. I'm more curious as to the other 8.2/8.3 changes that are proposed that AOL/Microsoft seems pretty big about.

I really fail to see any value in allowing transferring of blocks smaller than what is considered your minimum announce of a /24. If you aren't going to use a full /24, then rent the IP space from your datacenter.

 

[coreyman]

Not the best of ideas. No thanks. I'm more curious as to the other 8.2/8.3 changes that are proposed that AOL/Microsoft seems pretty big about.

I really fail to see any value in allowing transferring of blocks smaller than what is considered your minimum announce of a /24. If you aren't going to use a full /24, then rent the IP space from your datacenter.

Can you guys link this article please?

 

[rds100]

You think many providers would accept shorter than /24 prefixes? It would take between years and forever for some of them to adjust their filters.

Probably ipv6 would be implemented sooner.

@Francisco which L3 switches do you have in mind with 512K TCAM limit?

 

[Francisco]

You think many providers would accept shorter than /24 prefixes? It would take between years and forever for some of them to adjust their filters.

Probably ipv6 would be implemented sooner.

@Francisco which L3 switches do you have in mind with 512K TCAM limit?

The brocade RX's are like that and the Cisco 65XX's are too unless you get the highest spec addon card.

I know this is older gear, but it's still heavily used.

Francisco

 

[SPINIKR-RO]

I have seen this get brought up more than a few times. I doubt a vote would pass for such a thing.

 

[rds100]

If someone is running with a 512K TCAM limit today, he is already in trouble. The DFZ table today is between 480k and 490k and growing. It will only take some months before it exceeds 512K.

 

[SC-Daniel]

Microsoft hired a 10 year veteran from ARIN (David Huberman) to head their global IPv4 and IPv6 initiative or whatever it is they are calling it.

David has been extremely active in the PPML and ARIN-DISCUSS lists to try and shape policy.

When you have someone who used to enforce the policy for 10 years and saw how things were done and handled, then you take that person and place them into the role at a large corporation that then works that system from the other direction, of course you are going to start seeing large moves in the policy to fit that large corporations needs, while still appearing to be "for the greater good of the Internet"

The argument being made is that ARIN is just a database for number resources, and not the routing police, so they shouldn't govern saying that they won't offer less than /24 assignments or transfers for those who do not need global routeability.

What needs to happen to combat this is to start pushing for and lobbying RFCs to set a solid minimum announcement size into the global table.

-Daniel

 

[coreyman]

So what about IPV6, there are TONS of ips in all of the subnets (which there are way more of), and ISPs are generally given a /32 (correct me if i'm wrong). How can we not handle /24's coming into the routing table when we have 65,536 /64's in every /48 ??? Even if DC's don't route /64's there are still 65,536 /48's in every /32..... So that is an additional possibility of 4294967296 more subnets in the routing table right?

Scratch that - I'm stupid - there is no ARP table for IPV6

 

[SkylarM]

Scratch that - I'm stupid - there is no ARP table for IPV6

As a note, they aren't trying to make a transfer minimum a /24 -- that's what it is now. They want to scrap that ENTIRELY so people can transfer say /28's.

 

[Wintereise]

They can do whatever they want, I for one am not opening my filters for anything less than /24s.

Will simply send it to defroute if I don't have it in the table, and at that stage, it's the NSP's headache.