Re: Send-n, rat holes and real issues

Otmar Lendl wrote:

> Compare the # of domains out there with the number of routes in the
> global BGP routing table. Roughly, there are 3 orders of magnitudes 
> between the two.

Ok I think things are getting confused between routing and prefix
lengths here again. If route information is store there would be a lot
of data, if only prefix information is stored you store the general
cases then simply add exceptions as needed, the 1% Brian was pointing
out in an email the other day.

> Who announces a length for a prefix? The carrier (or enterprise) who
> operates the prefix. Given any leeway from the NRA, he is the only
> one who definitly knows.
> 
> Any porting operation leads to someone to announce the length of the
> number/prefix he just ported in. With open numbering plans, this ported
> number does not need to have the same length as before.
> 
> Thus, taking open numbering plans into account, the size of the
> BGP-like *whatever* table is O(#blocks + #ported numbers).

You are thinking about this wrong, the only time someone needs to
announce a prefix is if the length is different, countries with fixed
length dial plans would have very few entries.

Take Australia for example, I downloaded the entire file of allocated
numbers from the regulator a while back and ran a script across it.

http://www.e164.org/wiki/AU_Dialplan

For the entire country there is only 30 or 40 lines to describe the
entire prefix set 100% perfectly, this includes numbers that have been
ported, free phone numbers, so on an so forth.

Lets be clear here, we're only dealing with prefix lengths, and only
those that vary need to be stored as exceptions from the majority.

-- 

Best regards,
 Duane