[arin-ppml] Draft Policy 2010-2: /24 End User Minimum Assignment Unit - Correct Title
jim at tgasolutions.com
Mon Jan 25 15:32:50 EST 2010
>>> Secondly, if you can only justify a /23 then your actually better
>>>off "getting another ISP to advertise a competing ISP's space" than
>>>in attempting to get your own /23 or /24, even if this policy is
>>>approved. The reason why is that while your /23 or /24 would be
>>>a part of the DFZ, it would be so small in comparison to most of
>>>the blocks advertised on the Internet that there's a lot of networks
>>>(particularly running on routers with low ram in them) that will
>>>filter your /23 or /24 out entirely, and just default-route to
>>>one of their upstreams to cover the advertisements smaller than their
>>>filter break. It is a mistake to assume a /24 is globally visible
>>>in all routers on the Internet, if you spend some time working with
>>>different Looking Glasses (you can find a list on traceroute.org) you
>>>will quickly see that the smaller the advertisement the fewer routers
>>>it appears in. You may get suboptimal routing if you use a
>>>smaller block that is not part of a supernet, then if you use a
>>>small block that IS part of a supernet.
>>> Also, your going to quickly find that it's just as difficult to
>>>get a small portable block advertised from those "competing ISP's"
>>>as a small block that's assigned from an ISP.
Please forgive the intrusion... I post rarely---
As a consultant, I have several customers that are multi-homed. Many have the issue of getting
IP Space from an ISP. Some ISP's have to be read ARIN policy on the multi-home requirement
To allow for a /24.. Granted some of these customers cannot truly justify a /24 for that site.
HOWEVER-- there are concerns for them to do some other redundancy options..
Correct me if I am wrong-- But over on Cisco-NSP on Jared Mauch's site, I remember a conversation
A few months ago covering the /24 filtering.. many thought it was only being done on old routers..
And someone went so far to say that it was only done when it was a lower tier edge from a small ISP.
I think there was a branch on that thread that mentioned a default injection too.. But lots of water has
Flowed under the bridge since then....
As it is written today, I believe it sounds reasonable... and actually have a few end users that would benefit.
I am specifically addressing the small user, not the IBM as mentioned in the thread by Bill Herrin.
He has a very valid point, IMHO...
I know there are lots of folks here that members of the other RIR's lists.. so:
My parting shots are this:
What are the other RIR's doing?
IF they are already doing this, why is ARIN not?
IF they are not, what makes ARIN different?
This is not to say, ARIN should emulate the other RIR's but their logic may discount or account for issue or non-issues we may see here...
More information about the ARIN-PPML