[address-policy-wg] 2006-05 New Draft Document Published (PI Assignment Size)

On 21/10/2010 13:23, James Blessing wrote:
> How about the following situation:
> 
> I have 256 machines and 1 router, that's 257 addresses required. Under
> the new wording I can't then have a /23 because I have a requirement
> for 253 more addresses to make it up...
> 
> (I admit its unlikely but a potential situation that needs to be resolved)

So, can I take it then that you broadly agree with the principle of the
policy, but it's just the details which you believe need to be worked on?

There are edge cases which cause this policy to creak, and it's not hard to
construct "I Cannot Be Played on Record Player X" situations.  The policy
proposal is not intended to fix 100% of all situations.  Instead it is
intended to deal with ninety something percent of them.  If a policy were
created to deal with all potential situations, it would get messy.  I did
put some thought into this, but retreated from the idea.

Incidentally, the reason that 248 was chosen as a cut-off point was that
/29 is realistically the smallest chunk of address space that you would
want to run a multihomed site on.  /29 gives you space for your router
interface + two endpoint addresses on a LAN.

I already hear you saying "But, but, BUT!  You _could_ run a multihomed
site on 1, 2 or 4 addresses".  Of course you could.  But this is not an
exercise in playing number games.  The intent of /29 is to deal with the
smallest realistic network that an end-user is actually going to deploy
when they actually go multihomed (1 router + 2 hosts).

We could argue endlessly about 248 vs 252 vs 254.  But in practice, the
exact choice of number is going to make very little difference.

> There is also a potential issue where you have a requirement to subnet
> multiple locations where you are using a number less than the full
> number of addresses (eg 33 sites with /29 but only 5 address being
> used at each site to round up to a /23 you need more that 248
> addresses...)

This is already a problem with the existing assignment rules.  I don't plan
to fix this :-)

> "Further assignments under section 6.10 will not be permitted for an
> End User until all existing assignments have reached 80% utilisation"

I don't think it would help to complicate the proposal much more.

"Le mieux est l'ennemi du bien".

Nick