Re: [Idr] WGLC on draft-ietf-idr-as-private-reservation-00

[Jon Mitchell <[e-mail]>]

Inline...

On Sun, Dec 16, 2012 at 07:14:29PM +0000, Nick Hilliard wrote:
> 
> 1. "router bgp 65535" is quite common although it is officially a reserved asn.

Part of this (I believe) is due to the difference between RFC 1930 and
IANA reservation of original range, which is clarified by this draft.
Others are people not reading either, which is always likely...

> 
> 2. there are large spikes at the beginning of the range, around 65500, and
> right near the end of the private ASN range.
> 
> 3. there are smaller spikes at every asn which is evenly divisible by 100,
> and also at 65412 (i.e. 64512 with second two digits transposed - there are
> a lot of dyslexics in the industry), 65432, 64555, 65111.

> Speculating wildly, I suspect the underlying reason for this is that people
> find the numbers they've chosen to be easier to remember when working with
> them.  After all, we're human.

Nick - I assume you are willing to accept a modification to the proposal
that makes it 4200000000 - (end) even with your stated preference for a
smaller number of digits ?  This provides lot's of round numbers that
can be subdivided as folks see fit inside their own organizations.  I'd
point out that operators who plan on using only a small segment of the
new space are free to embed silliness or logic (depending on your point
of view) into the most significant x's like we have seen with IPv6, CLNS
addresses, and other large numbers if they are overly concerned about
typos or make it less likely to visually mis-identify a long string of
zeros.  I don't think we should carve out of the middle of the ASN space
for private use if we can avoid it nor think too small since we are
making a change and would prefer to progress this proposal rather than
debate the billions of options as I've stated before.

I'd prefer to minimize changes to the original doc at this point to only
those were there is strong concensus, that seems to be developing for a
change of the start of range to a decimal boundary (although I'd love to
hear from a few others on their preference between the two proposals on
the table).

Jon