Analysis of the ‘Maximum Length’ Option in Certification ROAs
[apologies for duplicates] Dear colleagues, We just published an article on RIPE Labs on the usage of the 'Maximum Length' option in ROAs. We compared all the prefixes in the Repository which are covered by a ROA to the prefixes seen by RIS. It gives a good indication how people are using this option, and what the effects are when people base routing decisions on validation results. You can read the article here: https://labs.ripe.net/Members/AlexBand/using-the-maximum-length-option-in-ro... Kind regards, Alex Band Product Manager
hi alex, one thing that interests me which i did not see in your analysis. or maybe i just need more coffee. how many, what proportion of, bgp announcements were for prefixes longer than the allocation in the roa and were properly described by a max-len? as to your choices, i would go with 1 or 2 (make it a mandatory blank field, forcing the user to make an explicit decision). 3 and 4 are horrible. randy
On 19 Apr 2011, at 21:40, Randy Bush wrote:
hi alex,
one thing that interests me which i did not see in your analysis. or maybe i just need more coffee.
how many, what proportion of, bgp announcements were for prefixes longer than the allocation in the roa and were properly described by a max-len?
Good question, let me look into that. Right now the analysis just focussed on the amount of prefixes that would be invalid because of incorrect usage of MaxLength, but it's quite likely a number of ROAs were created simply with /24 as MaxLength to allow for the freedom to deaggregate (or ease of use), when in reality the AS is only announcing a less specific aggregate.
as to your choices, i would go with 1 or 2 (make it a mandatory blank field, forcing the user to make an explicit decision). 3 and 4 are horrible.
Duly noted. -A
Alex, Randy, On Wed, 20 Apr 2011, Randy Bush wrote:
hi alex,
one thing that interests me which i did not see in your analysis. or maybe i just need more coffee.
how many, what proportion of, bgp announcements were for prefixes longer than the allocation in the roa and were properly described by a max-len?
as to your choices, i would go with 1 or 2 (make it a mandatory blank field, forcing the user to make an explicit decision). 3 and 4 are horrible.
Agree, or add an option 5; make a prediction (based on RIS) what the proper max-len should be. Jac -- Jac Kloots Network Services SURFnet bv
participants (3)
-
Alex Band -
Jac Kloots -
Randy Bush