Re: [Tm-rid] RAA/HDA - drawing the line in 32 bits

[Robert Moskowitz <[e-mail]>]

On 11/21/19 3:16 PM, Tero Kivinen wrote:
> Robert Moskowitz writes:
>> In draft-moskowitz-hip-hierarchical-hit-01.txt, I use 14 bits (16,384)
>> for the RAA space and 18 bits (262,144) for HDA within each RAA.  Seems
>> reasonable enough for me.  Good as any guess about the future.
>>
>> Tero Kivinen suggested a soft separation with the HDA in little endian
>> order.  This way we might have time to figure out where we need more,
>> RAAs or HDAs.
>>
>> Sounds interesting, but I see potential problem(s) if one RAA really
>> runs up its number of HDAs and thus limits the number of RAAs in total
>> as a result.
> I think we should be strict hard limit for RAAs for how many HDAs it
> can before it needs to come back for asking changing. For example say
> 12-bits for HDAs and if it looks like there are going to be multiple
> RAAs running out of 12-bit space, then we can expand it to 16-bit etc.
>
> If it just one RAA getting very large, we can of course give him
> 2nd RAA number.

Another item to toss into the mix is an HDA requesting a second or so 
number.  After registering so many HHITs, the collision rate becomes 
'unacceptable' and the move to a new HDA number for new registrations.

How many deliveries might UPS make a day via UAs?  A new HHIT per 
delivery or only per day?  This could run through a lot of HHITs.

Will some HDAs deprecate expired HHITs and allow reuse or keep the 
numbers around even after they are released by the owner?

All policy issues that will impact the hierarchy space.


> The IPv4 addresses just showed that we can never get the split right
> on the first time...
>
>> Oh, and the 32bit space will be represented in DNS as nibbles.

-- 
Standard Robert Moskowitz
Owner
HTT Consulting
C:248-219-2059
F:248-968-2824
E:[e-mail]

There's no limit to what can be accomplished if it doesn't matter who 
gets the credit