[c-nsp] Growing BGP tables

[David J. Hughes]

On 22/11/2004, at 5:54 PM, Gert Doering wrote:

> Surely it needs to be done with great care.

Very, very true.


> OTOH: people are free to renumber into aggregateable upstream space if
> their /24 stops working.  Why do I have to pay for their convenience?

And for those that obtained a portable /24 from AUNIC because they were 
small but multi-homed?  Those networks have been working just fine for 
the past 10 years because (a) they are addressed based on the original 
standard and (b) the space was allocated to them by the legitimate RIR 
of the time.  So we penalise them because there are many larger 
providers who have de-aggregated their blocks to death?

Australia is a little different.  The 'net uptake here was huge in the 
mid 90's.  Many, many organisations were allocated /24's and all looked 
good in the world because Geoff Huston continued to announce 203/8.  
Now that there are more transit providers things look less "orderly" as 
those smaller organisations announce their /24's and there's no 
distinct aggregate floating around.  However, I don't see this being a 
problem of "their convenience".  More a bi-product of people doing the 
right thing - at least as it was at that point in time.

Dropping /24's from the original class C space would, IMHO, be wrong 
from a standards point of view, and unnecessarily disruptive from a 
connectivity point of view.  That "great care" you mentioned could be 
applied to fix the table size without knowingly breaking things for 
people who "did the right thing", albeit it 10 years ago.



David
...