[arin-ppml] IPv6 Non-connected networks

[Michael Richardson]

>>>>> "michael" == michael dillon <michael.dillon at bt.com> writes:
    >> It seems like we're back to ULA again.  It seems that we all 
    >> want the same thing, "unroutable" address space for internal 
    >> use only.  There are any number of reasons to have RFC 
    >> 1918'ish space in v6 and the real difference now is that, 
    >> given the size of the v6 address space, we don't have to 
    >> allocate blocks to be used by all providers.  Instead, we're 
    >> in a position where we *can* give everyone their own block.
    >> 
    >> We could even do the old multicast thing and auto-assign a 
    >> ULA (or whatever it's called) block to each ARIN allocation.  
    >> In thinking about it, I do agree that having a defined block 
    >> for filtering purposes is preferred.

    michael> GLOP addressing uses the 16-bit ASN to specify a /24 block 
    michael> of multicast addresses, if I remember correctly. So, if we
    michael> used the 32-bit ASN to specify a /48 block of private use
    michael> addresses, we could probably fit it in. 

    michael> I don't believe that this has ever been proposed before
    michael> so it may be worthwhile writing a draft to see what happens.
    michael> And why not also include the "assigned ULA" addresses in 
    michael> the draft as well. If people like one idea better than the
    michael> other, then simplify the draft.

    michael> So we would have

    michael> 16 bits defined prefix for private use IPv6 addresses
    michael> 32 bits ASN number
    michael> 16 bits subnetting space
    michael> 64 bits IIDs
    michael> Total 128 bits of IPv6 address.

So this is basically a gift of a /48 to every owner of an ASN.
To most of them, they don't need an extra /48.  A few could use them,
but as John Santos says, most of the Enterprise types who need an
RFC1918 or swamp-class-C equivalent, they don't have an ASN already.

    michael> And given that there are some private-use ASNs defined, that
    michael> gives everyone the option to use those private-use addresses
    michael> whether or not they have an assigned ASN. That is essentially
    michael> the same as RFC 1918 addresses in IPv4.

The combination of the private-USE ASN with this proposal makes this
into EXACTLY rfc1918, aka "SITE-LOCAL" addresses.

This is a bad idea --- make it easy to get address space, and people
will use unique address space for unconnected things.  Your description
of COINs was very relevant.

So far I haven't seen a proposal (other than 103) which really addresses
this need.   It seems like the best answer is an RFC to be written to
tell IANA to delegate FC00::/16 to the RIRs.

-- 
]       He who is tired of Weird Al is tired of life!           |  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr at sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
   Kyoto Plus: watch the video <http://www.youtube.com/watch?v=kzx1ycLXQSE>
	               then sign the petition.