Re: More route-table bloat vs. ARIN micro-allocations

[I Am Not An Isp <patrick () ianai net>]

At 12:14 AM 2/21/00 -0500, jmalcolm () uraeus com wrote:

>701 per se is not the problem, as proportionally it's quite a bit
>better than most of the other large ASes. It only appears on the list
>because it is large in absolute numbers.

Perhaps 701 is not a problem, but I see at least three separate UUNET ASes 
on the CIDR report.


>Now 7046, which is UUNET customers announcing routes themselves, is
>pretty bad, and is a slightly harder problem to solve, although not
>impossible by any means.

AS7046 was given to each of those customers by UUNET.  It is used (at least 
in the cases I have seen) by networks with two or more connections to 
UUNET, for BGP load balancing.  Please note these are networks which have 
no other upstream, nor their own AS.  So, even though they have multiple 
links to UUNET, they are still "single-homed" as far as the rest of the 
'Net is concerned.

The overwhelming majority of prefixes in AS7046 are useless, as they are 
deaggregates of UUNET CIDR blocks, and not visible anywhere but through 
UUNET.  (There are some /16s which are not part of a larger CIDR sourced 
from AS701, so you cannot just filter AS7046 from your table without 
causing some disconnectivity.)

UUNET has had *over* a year (I notified them in November of 1998 - who 
knows how long they knew about it before that) to clean up/filter *one AS*, 
and has not done it.

Please note I tried to inform UUNET politely and privately about this, and 
waited quite a while for them to fix it.  After repeated attempts to get 
them to do something, a UUNET engineer told me I am not smart enough to 
understand how hard it is to filter an AS in a network as big as 
UUNET.  (Interesting way to treat a customers, wouldn't you say?)

As I am obviously not qualified to comment, the rest of you are welcome to 
draw your own conclusions on whether 16 months is enough time for even 
UUNET to filter one AS.


Another amusing part is that AS7046 shows up in the CIDR report every week 
(usually #2).  Then again, if AS7046 were counted as part of AS701 (which 
it essentially is), AS701 would regularly be #1 in the CIDR report.  (Maybe 
Tony should make a change to the report?)

Lastly, please note AS7046 is *growing*, not shrinking.


>>Yes, I'd rather see additional routes so that more providers are better
>>connected.  Otherwords Micro-Allocations from RIR's would provide:
>>
>>1. Better connectivity options for the bulk of the providers out there (not
>big ones)
>>2. Flexibility to add or change providers for the ISP's
>>3. Encouragement to multi-home by allowing MA's
>>4. Less hassles and negative impact on customers from renumbering.
>>   (NSI Makes it a real hassle to renumber via updates to host records)
>
>This is not compatible with the goal of holding steady the number of
>routes.

If these micro allocations were made to multi-homed networks, it would not 
grow the table, as multi-homed networks already have an AS and appear in 
the table.  It might even shrink the table as people with multiple 
non-aggregateable /24s were given, say, a /22.


>>But the widespread deagg to /24 doesn't seem to really help improve
>connectivity
>
>I suspect a lot of it is not intentional.

Does that make it okay?

TTFN,
patrick

--
  I Am Not An Isp - www.ianai.net
  ISPF, The Forum for ISPs by ISPs, <http://www.ispf.com>
  "Think of it as evolution in action." - Niven & Pournelle
  (Enable?  We dunt need no stinkin' enable!!)