Re: [lir-wg] AS Number Policy - continued
Not really important, but just to avoid being counted as agreeing :-)
Horrible thought? :)
No :-)
Anyway, I think that everyone have suggested to do both with a OR in between.
...no objection against the OR, but the result of the OR should not be the _only_ input to a decision process. It would be an efficient mechanism to be used _in support_ of continued AS# holdership, though.
Agreed. OTH, this discussion to me was for new AS assignments that are not yet used. Which is somewhat easier to deal with than with existing AS:es. AS286 beeing an example that is announced but the question is if we can consider it "as in use" ?
True. But to me it was not obvious yet, which "subset" of existing AS#s we are talking about, and whether the procedure would be pro-active only or post-factum as well.
Basically just claiming that "yes I do peer at backwater-IX with Farm-IP and Countryside Networks and I don't get transit from anywhere" should not be enouhg.
I disagree.
I read this as you mean that RIPE NCC shoudl trust the word of the AS number holder?
Yes. Please see below.
RIPE should be able to verify that this is true.
Some nit-picking...
RIPE (the community): I don't think so - other than having access to some publicly accessible documentation, e.g. in the Routing Registry.
My mistake. I ment the RIPE NCC.
I was reading it that way, too. But my experience with talking to people who for whatever reason get hold of these messages, and do not necessarily have the "implicit" mindset get mixed up easily.
The NCC: maybe, if and when we can agree on the criteria, and the cost for verification vs. the result.
Well, if we consider the RIPE db OR announced (or both - which is what it is supposed to be if the latter is true) it's not that hard. First, a requirement to register the AS number policy to keep it would be a easy task.
...and would actually, sort of through the backdoor, help to achieve better population of the Routing Registry. As you state, the quality of the data is a different issue, I agree. And defining the proper place for a registration is another "minor" technical issue (Q: portability of AS numbers amongst RIR service areas?). But we could give it a try...
Basically RIPE could then check assigned AS:es to registred. Still, the object does not have to be upto-date or actually reflecting anything.
Correct.
Second, as I belive there is so few assigned AS:es that never make it to the global routing tabele, I would like to define a few points of checks. These could even be route servers and this could be included in the automation. It could also be from the view of the test-traffic boxes.
I think we've been there: there is some chance that you cannot see those beasts from those places. Thus my claim (see above) that a statement of use (format to be discusssed) should be enough. For me it's a matter of the "10/90 rule": the cost of achieving 90% of your goal is 10%, doing the remainnig 10% is going to cost you 90%...
Best regards,
- kurtis -
Cheers, Wilfried.
Agreed. OTH, this discussion to me was for new AS assignments that are not yet used. Which is somewhat easier to deal with than with existing AS:es. AS286 beeing an example that is announced but the question is if we can consider it "as in use" ?
True. But to me it was not obvious yet, which "subset" of existing AS#s we are talking about, and whether the procedure would be pro-active only or post-factum as well.
Well, I think you have a valid point in that if we should start with this, we might as well do it post-factum as well. Problem then is that we need to automate this. And that will a task for the RIPE NCC, and require resources. So, I think for the distinction to make sense - we should first try and figure out how large the "current" problem is. See my previous posting, I think we need a comparison of the RIPE DB to a full BGP table. then we could pick a few AS:es and try and analys them.
The NCC: maybe, if and when we can agree on the criteria, and the cost for verification vs. the result.
Well, if we consider the RIPE db OR announced (or both - which is what it is supposed to be if the latter is true) it's not that hard. First, a requirement to register the AS number policy to keep it would be a easy task.
...and would actually, sort of through the backdoor, help to achieve better population of the Routing Registry. As you state, the quality of the data is a different issue, I agree.
Yes. Actually I would suggest that we would make this a requirement for keeping and acquireing an AS. This would require a policy change though I guess and some tools to verify. See above.
And defining the proper place for a registration is another "minor" technical issue (Q: portability of AS numbers amongst RIR service areas?). But we could give it a try...
Well, portably I guess is an issue anyway as the database refer to each other (or do they only do that if there is no match? I am not really into the whois data...).
Second, as I belive there is so few assigned AS:es that never make it to the global routing tabele, I would like to define a few points of checks. These could even be route servers and this could be included in the automation. It could also be from the view of the test-traffic boxes.
I think we've been there: there is some chance that you cannot see those beasts from those places. Thus my claim (see above) that a statement of use (format to be discusssed) should be enough.
If we take a few steps back in the discussion - we said (well at least I did..:) ) that a ISP would be contacted after [a] time and asked if they had used the AS. If not they would be given [b] time (unless they had come to the conclusion they did not needed it or could meet the timeline) to correct this. I would suggest that after [a] time they are required to make a statement along the suggested lines of : - Upstream - Peers - etc. Perhaps as with the current IP space useage form (after just having filled in one and trying to get it approved I am sure that is enough to scare most people away anyway...:) ). However, I suggest that then, after [b] time they need to prove that this is now in use. Why? Well, if we assume that [a] is six months and [b] is six months, they have had a year. In that time they should have got whatever they needed the AS for up and running.
For me it's a matter of the "10/90 rule": the cost of achieving 90% of your goal is 10%, doing the remainnig 10% is going to cost you 90%...
Agreed, but it is also a question of how the cost really is. Compared to work we (well, the RIPE NCC) spend on IP space, some mechanisms here shouldn't matter to much... - kurtis -
participants (2)
-
Kurt Erik Lindqvist -
Wilfried Woeber, UniVie/ACOnet