[ppml] /48 vs /32 micro allocations

Owen DeLong owen at delong.com
Tue Mar 15 14:33:24 EST 2005



--On Tuesday, March 15, 2005 12:15 PM -0500 Kevin Loch <kloch at hotnic.net> 
wrote:

> bmanning at vacation.karoshi.com wrote:
>> 		) there will be enough IPv6 space to last for the next
>> 		  30 years
>
> Increasing the minimum allocation size for a few hundred micro
> allocations to 1 / 4billionth of the total address space is not
> wasteful. If you are concerned with relative waste you might want
> to look at the HD ratio of 0.8 (!) that is resulting in /19 allocations
> (that's 8192 /32's right there).
>
Do you have any evidence to support that the number of microallocations
would remain so small?  I tend to expect that it would not.

>> 		) no one would -ever- dare treat IPv6 space as anything
>> 		  useful in less than /32 chunks
>
> This isn't about /33-/47 it's about /48 (see below).
>
In that case, why aren't you asking to shift /48s to /47s instead of
/32s?  What's the point of addressing /48s without addressing /33-/47
if you're going to move /48s all the way to /32?  That seems very
illogical to me.


>> 		) no harm will insue if folks advertise routes to space
>> 	    	  that has no endsystem attached.
>
> Others have addressed authentication but I think maintaining the MRU
> above /48 would help. The larger prefix an abuser has to announce the
> more likely it is to be noticed.  The presence of end systems doesn't
> help at all in IPv6 (host density is impossibly low) though I can see
> problems if the micro allocation is larger than the effective MRU
> (abuser announces more specifics). There's nothing to stop you from
> announcing the more specifics yourself, at least then the larger RIR
> allocation won't encourage looser filters.  If /32 micro allocations are
> so harmful, then why is that the policy of the other RIR's?  K, I and M
> root all have /32's allocated.
>
If you advertise the more specifics yourself, then, you have guaranteed
the problem you were claiming this was an attempt to avoid.  You can't
have your cake and eat it too.  Agreed that IPv6 is hopelessly broken
when it comes to allocation density.  However, making it 65,536 times
less dense will not improve the situation.

>> 		) we understand all there is to know about how applications
>> 		  and services will develop over the next two decades to
>> 		  presume a steady, consistant "burn-rate" of space.
>
> Right now we are talking about a few hundred micro allocations.  When
> (not if) we start talking about end site allocations we could consider
> smaller sizes than /32 but even then not /48 unless the intent is that
> all /48's (directly assigned or not) be globally routable.
>
There is already an end-site policy proposal in ARIN.  Thus, we _ARE_
already talking about end-site allocations.  Next.

>> 		) no fundamental technological changes will occur on our
>> 		  watch and if it does, it can be solved by new talent.
>> 		) that the IETF leaves routing development alone.
>
> Allocation policy is not set in stone.  RIR's can and do change policy
> as conditions change.  What is the right thing to do right now?
>
Ignore IPv6 as it isn't quite ready for the general public yet.  A number
of problems IPv6 needs to solve to be really worth migration are not yet
solved:

	Address portability or completely easy renumbering
		+	Including ACLs
			*	ACLs within the organization
			*	Related ACLs outside the organization which
				may or may not be known to the organization
		+	Including statically configured end hosts
		etc.

	How to handle a routing table of the size likely when IPv6 does
	become popular and NAT is no longer an issue.

	Mobile IP

	Route Authentication/Authorization

	Prefix Authentication/Authorization

To name but a few.


>>
>> 	If this is the basis of your argument, then i'll agree that its
>> 	a good proposal, if all my points are accepted and ISPs agree to
>> 	never announce anything smaller than what there RIR gives them.
>
> Of course they will.  I'm not suggesting we could maintain an MRU of
> /32.  With the quantumn nature of current address quidelines, there is
> all the difference in the world between an MRU of /48 and /47.
>
ISPs rarely agree on anything, and, the MRU is no exception.  ISPs do
one of two things:

	1.	What their main techie thinks is the right thing because
		he has the ability to set the policy on the routers and
		management/sales haven't noticed yet.

	or
	2.	What management/sales have dictated will be the policy on
		the routers because they see revenue behind it.

If there's revenue in advertising somebody's more specific, it will get
advertised.

> Would you agree that:
>
> - There is a benefit to increasing micro allocations above /48 WRT MRU

No.

> - There is some value between /32 and /47 where the benefits outweigh
> the risks?
>
No.

Owen


-- 
If it wasn't crypto-signed, it probably didn't come from me.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 186 bytes
Desc: not available
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20050315/e53a58c8/attachment-0001.sig>


More information about the ARIN-PPML mailing list