> On Feb 24, 2015, at 12:38 , Dean Pemberton <[email protected]> wrote: > > On Wed, Feb 25, 2015 at 6:20 AM, Owen DeLong <[email protected]> wrote: >>> Firstly I agree with Randy here. If you're not multi-homed then your >>> routing policy can not be 'unique' from your single upstream. You may wish >>> it was, but you have no way to enforce this. >> >> This is not true. >> >> You can be single homed to a single upstream, but, have other peering >> relationships with other non-upstream ASNs which are also not down-stream. >> These relationships may not be sufficiently visible to convince APNIC that >> one is multihomed, even though technically it is a multihomed situation. >> > > I don't agree (Damn and we were getting on so well this year =) ). > > I would argue that the situation you describe above DOES constitute > multihoming.
I agree, but it may not necessarily constitute “multihoming” in a manner that is recognized or accepted by the RIR. Clarification from APNIC staff on the exact behavior from APNIC could render this moot. However, I have past experience where RIRs have rejected peerings with related entities and/or private ASNs of third parties as not constituting valid “multihoming” whereupon I had to resort to “a unique routing policy”. > If an LIR were connected to an upstream ISP but also wanted to > participate at an IXP I would consider them to be multihomed and > covered under existing APNIC policy. What if they only connected to the IXP with a single connection? I’ve also encountered situations where this is considered “not multihomed” and to be a “unique routing policy”. > > I couldn't find the strict definition on the APNIC pages as to what > the hostmasters considered multihoming to be, but if one of them can > point us to it then it might help. Agreed. > > >> While I oppose that (and thus completely oppose the other proposal), as >> stated above, I think there are legitimate reasons to allow ASN issuance in >> some cases for organizations that may not meet the multi-homing requirement >> from an APNIC perspective. > > I really want to find out what those multi-homing requirements are. I > suspect that they amount to "BGP connections to two or more other > ASNs" > In which case I think we can go back to agreeing. As long as it’s not more specific than that (for example, two or more public ASNs or via distinct circuits, etc.). > > >> I think it is more a case that smaller and simpler policy proposals that >> seek to change a single aspect of policy are more likely to succeed or fail >> on their merits, where large complex omnibus proposals have a substantial >> history of failing on community misunderstanding or general avoidance of >> complexity. > > I can see your point, but taking a smaller simpler approach is only > valid once you have agreed on the larger more strategic direction. I > don't believe that we have had those conversations. I find that in general, the larger the group you are attempting to discuss strategy with, the smaller the chunks necessary for a useful outcome. YMMV. > We are seeing small proposals purporting to talk about multihoming, > but what they are in essence talking about is the much larger topic of > the removal of demonstrated need (as Aftab's clarification in the > other thread confirms beyond doubt.) Upon which clarification, you will notice that I switched to outright opposition to that policy. Frankly, you caught a subtlety in the language that I missed where I interpreted the proposal to still require justified need rather than mere announcement, but a careful re-read and the subsequent clarification of intent made it clear that I had erred. Further, note that I have always opposed this proposal as written, but offered as an alternative a much smaller change which I felt met the intent stated by the proposer without the radical consequences you and I both seem to agree are undesirable. > There is danger in the death by a thousand cuts. Many times you can't > see the unintended consequences until you are already down the track > of smaller simpler policy changes. I really don’t think that is a risk in this case. > As we are in Japan I offer a haiku: > > A frog in water > doesn’t feel it boil in time. > Do not be that frog. > > (http://en.wikipedia.org/wiki/Boiling_frog) I wish I could be at the meeting, but, alas, I’m here in the US looking on from afar. Owen * sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list [email protected] http://mailman.apnic.net/mailman/listinfo/sig-policy
