Andy Linton sent me this response, which he tells me he intended for nznog as well. Below is Andy's response, and then my reply to Andy's response. Both are at least as relevant to the list as discussing Klez all day :-) Andy's response: On Wed, 11 Sep 2002, Ewen McNeill wrote:
In message
, "Arron Scott" writes: I get endless number of customers who want to peer with more than one ISP at a time, usually with the intention of resiliency at both the institutional and upstream provider levels. [... Want Class C, public AS, advertised prefix... ] Now ... what I was thinking is ... can we do this without the rare (and increasingly difficult to obtain) Public AS numbers.
There are proposals to increase the size of the AS numbers from 16 to 32 bits but that soesn't solve the immediate difficulties in getting an AS number.
I'm also in touch with clients that would like to do this -- they don't especially care about the Class C, or Public AS, etc, but they would like some resiliency for any disconnections from their provider or routing issues for their provider, often because they have clients that have that as a tick-box feature.
I've heard some very interesting discussions about flap damping at the RIPE meeting this week. See http://psg.com/~randy/020910.zmao-flap.pdf. The important thing to note is that the thing that is damped is prefixes so dual homing may not give you all you hope for if flap damping is in use.
Could we have a publically agreed on pool of Private AS numbers that enterprises can use to peer with service providers. The pool would administered by a "impartial" group (maybe WIX/APE). The AS number would then be stripped by both higher-order ISPs and and the IP address potentially unsuppressed by the ISP who owns the IP address aggregate.
If I'm not mistaken about what you mean, this is what WIX (the route reflectors) are trying to do.
Dean Pemberton and I have been talking with Simon about trying to get WIX and APE whois servers running so that people could start registering routes and AS policy for RPSL use in NZ. Using these to allocate locally scoped private ASes for use on the WIX and APE would be a doddle.
The WIX approach seems to work, but: - it appears some providers do not peer with the WIX route reflectors, or do so only to advertise their routes, not to allow in extra routes
- most providers filter "long" prefixes (and most enterprises without public ASes tend to have only long prefixes). Certainly it's my impression that pretty much all providers filter anything longer than /24, and some appear to even filter /24, /23, /22, etc, meaning the prefixes advertised by enterprises are likely to be filtered.
It seems difficult to believe that the relatively small additional number of prefixes being announced by the enterprises on the WIX is significant. The total number of prefixes is around the 250 mark and many of those come from the "providers" anyway. Filtering out the few dozen prefixes that are /25 or greater would seem to have little impact on table size and the kosher ones could easily handled in a prefix-list.
So from an enterprise point of view WIX (the route reflectors) are useful for picking up direct routes to providers POPs, but this tends to lead to asymetric routes to/from the providers (and hence no real resiliency).
So it would be nice to get the providers to use them. Some of them resist but surely this is an issue that could be driven by their respective customers. E.g. "As my provider I need you to provide a resilient efficient local mesh at WIX and APE. Perhaps I should consider moving to a provider who...."
And it's useful for enterprise to enterprise peering (indeed one of my clients connected to the WIX route reflectors soley for the enterprise to enterprise peering; at the request of the other enterprise).
The WIX route reflectors (according to the route table on a router peering with them) appears to have about 50 systems peering with it at present.
But without providers willing to (a) peer with the route reflectors, and (b) accept long prefixes (at least locally), it's probably not going to be of general use for resiliency.
To be of much general use providers would have to be willing to at least accept /24 prefixes locally, even if they're part of a larger supernet advertised by another provider. Longer prefixes would be desirable, because otherwise anyone wanting resilency with CIDR space would need to beg for a class-C sized CIDR chunk.
See above
---------------------------------------------------------------------------
And my response to Andy:
In message
On Wed, 11 Sep 2002, Ewen McNeill wrote:
I'm also in touch with clients that would like to do this [advertise their blocks to multiple providers for resiliency]
I've heard some very interesting discussions about flap damping at the RIPE meeting this week. [...] The important thing to note is that the thing that is damped is prefixes so dual homing may not give you all you hope for if flap damping is in use.
For both the clients I have in mind most of their benefits would be gained in about the first 5 hops or so (ie, within New Zealand), as that is the main market they're serving. So how important flap damping was to that would largely depend on how soon it was applied.
Dean Pemberton and I have been talking with Simon about trying to get WIX and APE whois servers running so that people could start registering routes and AS policy for RPSL use in NZ. Using these to allocate locally scoped private ASes for use on the WIX and APE would be a doddle.
This would definitely be a good thing. RPSL (from your talk at the Uniforum conference) looks like a Good Thing (tm), and a whole lot more sensible than the alternatives (yet more prefix listson nznog!). Scoped private ASes would also solve the immediate problem -- probably for Aaron too (hence my suggestion you mention this on nznog, and/or directly to Aaron).
It seems difficult to believe that the relatively small additional number of prefixes being announced by the enterprises on the WIX is significant.
WIX is worth about 300 prefixes or so on an average day, and "generally" includes the POPs of most NZ providers. For an organisation hosting a lot of content, there's definitely a win for them in sending traffic "directly" to the POPs (and bypassing the ticket counter at their ISP :-) ). If those routes were symmetric, then there'd also be a resiliency win.
The total number of prefixes is around the 250 mark and many of those come from the "providers" anyway. Filtering out the few dozen prefixes that are /25 or greater would seem to have little impact on table size and the kosher ones could easily handled in a prefix-list.
Last I spoke to Simon about it, he was starting to put filters in on the expected prefixes from each peer with the WIX route reflectors -- which would make it possible to have a pretty "exact" set of expected prefixes. There are 27 prefixes in WIX (right now) longer than /24. Some of them are as long as /30s, and there's a bunch of /29s. One of the clients I have in mind has a /27 CIDR block. They're starting to run out of public address space (despite using RFC1918 everywhere it can be used, and lots of NAT), and will probably try asking for more space. But even that more space is likely to be only another /28, another /27, or maybe if they're very lucky one /26. My point here is that while there might not be very many prefixes longer than /25 in WIX now, that's sort of a self-fulfilling prohpehcy, because prefixes longer than /24 (or /23, /22, /21, etc -- I haven't managed to find out what's considered "best practice" by NZ providers) are being filtered out already by lots of people. So "what's the point?". If the answer is "everyone use RPSL, and persuade people to accept all properly described RPSL lists" then I'm all for it. If the answer is "only providers get to do peering", then that's tantamount to "only people with public ASes get to do peering". And may well lead to a bunch of people chasing public ASes (and provider independant space for that matter) when they don't otherwise need them. Ewen - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wednesday, September 11, 2002, at 04:19 PM, Ewen McNeill wrote:
Andy's response:
[...]
Dean Pemberton and I have been talking with Simon about trying to get WIX and APE whois servers running so that people could start registering routes and AS policy for RPSL use in NZ.
Just out of interest, what's the benefit in running yet another RPSL database?
Using these to allocate locally scoped private ASes for use on the WIX and APE would be a doddle.
Run and hide! There's no reason to break RFC1930. That makes things difficult and confusing, whereas obtaining a globally-unique ASN is simple and easy.
[...]
So it would be nice to get the providers to use them. Some of them resist but surely this is an issue that could be driven by their respective customers. E.g. "As my provider I need you to provide a resilient efficient local mesh at WIX and APE. Perhaps I should consider moving to a provider who...."
... is willing to surrender control of her routing policy to a best-effort coordination service with no responsibility for the quality of the routing data sent to or from her network?
Ewen says:
One of the clients I have in mind has a /27 CIDR block. They're starting to run out of public address space (despite using RFC1918 everywhere it can be used, and lots of NAT), and will probably try asking for more space. But even that more space is likely to be only another /28, another /27, or maybe if they're very lucky one /26.
Tell your client that a requirement to multi-home (whether to multiple transit providers, or to a single transit providers and multiple peers) is adequate justification for being allocated a /24 netblock from their transit provider. Ask, and it will be given.
If the answer is "everyone use RPSL, and persuade people to accept all properly described RPSL lists" then I'm all for it. If the answer is "only providers get to do peering", then that's tantamount to "only people with public ASes get to do peering". And may well lead to a bunch of people chasing public ASes (and provider independant space for that matter) when they don't otherwise need them.
If you want to multi-home using BGP, and you don't want to violate RFC1930, you need a globally-unique ASN. ASNs are not just allocated to providers. So, only people with public ASNs get to do peering, but that doesn't mean that only providers get to do peering. None of this is new. In fact, there was enough of this going on when I was still involved in AS4768 that I *documented* it: http://www.clear.net.nz/documentation/dedicated/multi.html Some of those documents are old enough that they have NZIX in their diagrams :) Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Dean Pemberton and I have been talking with Simon about trying to get WIX and APE whois servers running so that people could start registering routes and AS policy for RPSL use in NZ.
Just out of interest, what's the benefit in running yet another RPSL database?
Infact, it might be a very good idea to host an RPSL database in New Zealand, and ask the nice folks at Merit to mirror it. If only for such reasons as having nice fast access to the DB, and having it run by someone you trust. Moebius Systems is more than happy to run said irrd server in NZ on one of our servers if that's what needs to happen to organise the routing in NZ a little more. WRT private AS', I am not sure if this is a good idea, however I am also not sure that asking enterprises to get their own globally unique AS number is a good (or responsible) thing to do either. Perhaps we can form a "working group" (not the InternetNZ sense, the NZNOG sense, where we all go out for curry and beer) to discuss a standard for this sort of thing within NZ. I know this sounds yuck, but what about carefully controlled IGP's or maybe funky use of confederations? There is also one more way of acquiring /24 address space that hasn't been mentioned. We acquired the remnants of a company with one :) -- Cheers. James Tyson --- Samizdat New Media Solutions
On 12 Sep 2002, James Tyson wrote:
form a "working group" (not the InternetNZ sense, the NZNOG sense, where we all go out for curry and beer) to discuss a standard for this sort of thing within NZ. I know this sounds yuck, but what about carefully So how many are showing up for Curry and beer tonight?
Tonight in Auckland: http://auckland.thursdaynightcurry.com/ And other cities info can be found at Wellington, Melbourne, Sydney, and San Francisco . Thought not quite sure if convo at other cities apart from auck and wgtn would be fruitful. grin Lin - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On 12 Sep 2002, James Tyson wrote:
Dean Pemberton and I have been talking with Simon about trying to get WIX and APE whois servers running so that people could start registering routes and AS policy for RPSL use in NZ.
Just out of interest, what's the benefit in running yet another RPSL database?
Infact, it might be a very good idea to host an RPSL database in New Zealand, and ask the nice folks at Merit to mirror it.
Let me nail my colours to the mast. My preference is not to run a local database. APNIC has recently made the transition to RIPE V3 code and have highly available and fast servers in Brisbane. This data is mirrored at RIPE. The servers in Brisbane are around 90ms away. They will have full support for the relevant RPSL objects in the next few months - the database already supports these - hostmaster training is currently in progress. Dean and I have only discussed running a local service because we were concerned that local resistance to using a remote service might slow progress. IFF we run a database it would be with the goal of using it as a transition step.
If only for such reasons as having nice fast access to the DB, and having it run by someone you trust.
See above - I think this access is fast enough and I think APNIC can be trusted. They're also roughly in the same time zone and people are in general customers who have paid for this service.
Moebius Systems is more than happy to run said irrd server in NZ on one of our servers if that's what needs to happen to organise the routing in NZ a little more.
I'd recommend using the RIPE code as the authentication and database consistency mechanisms and search capabilites are much better.
WRT private AS', I am not sure if this is a good idea, however I am also not sure that asking enterprises to get their own globally unique AS number is a good (or responsible) thing to do either. Perhaps we can form a "working group" (not the InternetNZ sense, the NZNOG sense, where we all go out for curry and beer) to discuss a standard for this sort of thing within NZ. I know this sounds yuck, but what about carefully controlled IGP's or maybe funky use of confederations?
And here's the only time I'd advocate running a private database. If people want to allocate private ASes for use on a "private network" like Citylink (remember the 202.7.0.0/23 is not routed globally). In that case it wouldn't be appropriate to have Merit/APNIC etc mirror this although duplicate local servers could be run using NRTM.
There is also one more way of acquiring /24 address space that hasn't been mentioned. We acquired the remnants of a company with one :)
Mmm. - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wed, Sep 11, 2002 at 05:10:39PM -0400, Joe Abley said:
Just out of interest, what's the benefit in running yet another RPSL database?
WIX has 50+ private ASN peers, AFAIK you can't put info about private ASN into the public RPSL services, so if you want the value RPSL provides for private ASN (and I ohh so do), you run your own database.
efficient local mesh at WIX and APE. Perhaps I should consider moving to a provider who...."
... is willing to surrender control of her routing policy to a best-effort coordination service with no responsibility for the quality of the routing data sent to or from her network?
Ewen says:
One of the clients I have in mind has a /27 CIDR block. They're starting to run out of public address space (despite using RFC1918 everywhere it can be used, and lots of NAT), and will probably try asking for more space. But even that more space is likely to be only another /28, another /27, or maybe if they're very lucky one /26.
Tell your client that a requirement to multi-home (whether to multiple transit providers, or to a single transit providers and multiple peers) is adequate justification for being allocated a /24 netblock from their transit provider. Ask, and it will be given.
If the answer is "everyone use RPSL, and persuade people to accept all properly described RPSL lists" then I'm all for it. If the answer is "only providers get to do peering", then that's tantamount to "only people with public ASes get to do peering". And may well lead to a bunch of people chasing public ASes (and provider independant space for that matter) when they don't otherwise need them.
If you want to multi-home using BGP, and you don't want to violate RFC1930, you need a globally-unique ASN. ASNs are not just allocated to providers. So, only people with public ASNs get to do peering, but that doesn't mean that only providers get to do peering.
None of this is new. In fact, there was enough of this going on when I was still involved in AS4768 that I *documented* it:
http://www.clear.net.nz/documentation/dedicated/multi.html
Some of those documents are old enough that they have NZIX in their diagrams :)
Joe
- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
Oh for crying out loud. Four years ago, we started agressively filtering incoming advertisments on WIX from new peers. Three years ago, we had filter lists in place for *every* *single* *peer*. New peers are independantly vetted (I check with their upstreams that what they want to advertise is plausible). We are no longer just "best efforts", and have not been for many years, conversely, we haven't had a routing catastrophe for years either. I've pointed this out several times on NZNOG, and yet you continue to assert that it's not the case. <sigh> Cheers Si - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Thu, Sep 12, 2002 at 10:02:10AM +1200, Simon Blake wrote:
On Wed, Sep 11, 2002 at 05:10:39PM -0400, Joe Abley said:
Just out of interest, what's the benefit in running yet another RPSL database?
WIX has 50+ private ASN peers, AFAIK you can't put info about private ASN into the public RPSL services, so if you want the value RPSL provides for private ASN (and I ohh so do), you run your own database.
... or you use globally-unique ASNs. Using private ASNs for non-private applications is surely broken, especially when it's so trivial to obtain a globally-unique one. You might argue that it's inefficient use of a finite resource for enterprises (in the AS1918 sense) that are not transit providers to be allocated globally-unique ASNs, and I might agree with you. That's not a problem that's going to be solved in New Zealand, though; that's a problem that will be managed by IANA and the RIRs with allocation policies until someone comes up with a multi-homing and routing system that scales better than the one we have.
efficient local mesh at WIX and APE. Perhaps I should consider moving to a provider who...."
... is willing to surrender control of her routing policy to a best-effort coordination service with no responsibility for the quality of the routing data sent to or from her network?
Oh for crying out loud.
I'm not talking about ingress filtering done by the WIX route servers. By "responsibility for the quality", I mean: + having a route propagation path which is different to the packet forwarding path, which is a general problem of route servers on non-trivial layer-2 exchange fabrics; + having no contract/support relationship/whatever between operators connected to the route server, which is a general problem of multi-lateral peering. As to the "best-effort" bit, I thought you were; sorry if that's not the case.
I've pointed this out several times on NZNOG, and yet you continue to assert that it's not the case.
Nope. I don't remember making any comments ever about what ingress policy you were using for the route servers. I do remember making comments about it being generally hard for *other* people to come up with a sensible ingress policy for their session to the route server, though, which is quite different. Lots of people find your route servers useful, and that's great. They *are* useful. I was objecting to the idea that operators who don't use the route servers must be bad or stupid, or be otherwise unworthy of attracting customers, because I don't think that's reasonable; there are arguments for not using them, just as there are arguments *for* using them. Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wed, Sep 11, 2002 at 06:48:08PM -0400, Joe Abley said:
On Thu, Sep 12, 2002 at 10:02:10AM +1200, Simon Blake wrote:
On Wed, Sep 11, 2002 at 05:10:39PM -0400, Joe Abley said:
Just out of interest, what's the benefit in running yet another RPSL database?
WIX has 50+ private ASN peers, AFAIK you can't put info about private ASN into the public RPSL services, so if you want the value RPSL provides for private ASN (and I ohh so do), you run your own database.
... or you use globally-unique ASNs. Using private ASNs for non-private applications is surely broken, especially when it's so trivial to obtain a globally-unique one.
From my perspective, it *is* private usage - I want RPSL to manage the peers of the route servers on APE and WIX. If it gets used in a wider sense by the NZ Internet, then that's well and good.
You might argue that it's inefficient use of a finite resource for enterprises (in the AS1918 sense) that are not transit providers to be allocated globally-unique ASNs, and I might agree with you. That's not a problem that's going to be solved in New Zealand, though; that's a problem that will be managed by IANA and the RIRs with allocation policies until someone comes up with a multi-homing and routing system that scales better than the one we have.
Surely. But we all work with the tools we have, and there is a perception that global ASN are hard to obtain. I don't know if that's true (I haven't asked for one for years), but the perception is still there.
I'm not talking about ingress filtering done by the WIX route servers. By "responsibility for the quality", I mean:
+ having a route propagation path which is different to the packet forwarding path, which is a general problem of route servers on non-trivial layer-2 exchange fabrics;
Sure, but that's an equally good arguement for fixing the exchange fabric, since if you're under some kind of partial reachability cloud, then you're not getting full value for your connection anyway. WIX has never had such problems (presumably because its L2 fabric is primarily under the control of a single administrator). I have seen it happen on APE a couple of times though, so I agree that it is an issue. Eitherway, I think it's a compelling argument to fix the fabric, rather than abandon route servers.
+ having no contract/support relationship/whatever between operators connected to the route server, which is a general problem of multi-lateral peering.
Sure. We have various contracts available if people want to look at them, but generally, people don't. Funny, that :-). We're not seeking to supplant bilateral peering with the route servers, and we never will. What I am seeking to achieve is "lowest common denominator" peering where every entity attached to WIX/APE can advertise routes that it regards as "free", and listen to the "free" routes from other peers. Every entity can define "free" in whatever way it likes, and "none of my routes", "all of my routes" and "local Wellington or Auckland routes" are all valid definitions.
Nope. I don't remember making any comments ever about what ingress policy you were using for the route servers. I do remember making comments about it being generally hard for *other* people to come up with a sensible ingress policy for their session to the route server, though, which is quite different.
Sure, and I believe the RPSL stuff we've got planned with Andy will go a long way to helping peers make sensible ingress policy for their route server sessions.
I was objecting to the idea that operators who don't use the route servers must be bad or stupid, or be otherwise unworthy of attracting customers, because I don't think that's reasonable; there are arguments for not using them, just as there are arguments *for* using them.
Undoubtedly. I just think the arguements for peering with the route servers are significantly stronger than the arguements for not peering with the route servers, but then I'm biased :-). Cheers Si - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wednesday, September 11, 2002, at 07:39 PM, Simon Blake wrote:
+ having a route propagation path which is different to the packet forwarding path, which is a general problem of route servers on non-trivial layer-2 exchange fabrics;
Sure, but that's an equally good arguement for fixing the exchange fabric, since if you're under some kind of partial reachability cloud, then you're not getting full value for your connection anyway.
Right. The issue arises when there is a functional layer-2 path from the route-server to your router, but no functional layer-2 path from your router to the next-hop address for particular routes you learn from the route server. In those circumstances you will black-hole traffic across the exchange instead of losing the route and sending the traffic by some other path (which is what would happen if you only learnt routes directly from the other operator's router, with no route server involved). It's a failure mode thing, not a value for money thing.
+ having no contract/support relationship/whatever between operators connected to the route server, which is a general problem of multi-lateral peering.
Sure. We have various contracts available if people want to look at them, but generally, people don't. Funny, that :-).
There is no contract between ISP A and ISP B who receive each others' routes from the route server, though. That's what I was getting at. That's important to some people, and not just for lawyers-run-the-company, my-hair-is-pointy reasons.
I was objecting to the idea that operators who don't use the route servers must be bad or stupid, or be otherwise unworthy of attracting customers, because I don't think that's reasonable; there are arguments for not using them, just as there are arguments *for* using them.
Undoubtedly. I just think the arguements for peering with the route servers are significantly stronger than the arguements for not peering with the route servers, but then I'm biased :-).
I am not anti-route-server, on APE or WIX or elsewhere; I just don't think they have universal application. I have set up networks that use the APE route server, and I have also helped set policy in other networks which involved deliberately not connecting to the route server. I ordered the ASNs for the APE and WIX route servers, back in the day. I wouldn't have necessarily bothered if I thought they were a bad idea :) Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wed, Sep 11, 2002 at 09:16:38PM -0400, Joe Abley said:
On Wednesday, September 11, 2002, at 07:39 PM, Simon Blake wrote:
+ having a route propagation path which is different to the packet forwarding path, which is a general problem of route servers on non-trivial layer-2 exchange fabrics;
Sure, but that's an equally good arguement for fixing the exchange fabric, since if you're under some kind of partial reachability cloud, then you're not getting full value for your connection anyway.
Right.
The issue arises when there is a functional layer-2 path from the route-server to your router, but no functional layer-2 path from your router to the next-hop address for particular routes you learn from the route server.
In those circumstances you will black-hole traffic across the exchange instead of losing the route and sending the traffic by some other path (which is what would happen if you only learnt routes directly from the other operator's router, with no route server involved).
It's a failure mode thing, not a value for money thing.
Well, that's debatable - it's still broken, you're not getting value for money, and if the L2 connection over the exchange between the two ISP's is the only path between them, then you're still screwed. I think my observation still stands - if your L2 infrastructure is broken such that you don't have reachability between points A and B on that infrastructure, then that is the real problem, and that's what has to be fixed. What we're really discussing is whether the advantages the route servers bring in normal operation is outweighed by the disadvantages when the L2 exchange goes pear shaped. I think the advantages are greater (because experience tells me that the NZ L2 exchanges are reliable), you think the disadvantages are greater (presumably coz you've seen exchanges elsewhere go mental). I don't think you can broad brush the arguement either way.
There is no contract between ISP A and ISP B who receive each others' routes from the route server, though. That's what I was getting at. That's important to some people, and not just for lawyers-run-the-company, my-hair-is-pointy reasons.
Again, I'm interested in the local, specific cases, rather than in general contract theory :-). I fully agree that some peers may want to have contracts between each other, and the associated bilateral peering, and I've no problem with that whatsoever. That doesn't mean, however, that route servers are immediately removed from the discussion, it just means ISP's have different policies for routes learnt from the route server vs from bilateral peers. It is, as always, a value thing - you're getting a bunch of routes with little effort from the route servers - how much value do you get from that, vs the effort involved with initiating bilateral contractural arrangements with specific providers. How much do you trust the routes from each source? How important do you rank the need for legal comeback if stuff goes wonky? Each provider gets to make their own judgements on these things.
I ordered the ASNs for the APE and WIX route servers, back in the day.
Not strictly - I obtained the WIX ASN by mine own fair hand, it was the APE one you obtained (and for which we are eternally grateful, since it doesn't cost us anything :-).
I wouldn't have necessarily bothered if I thought they were a bad idea
Nor I :-). Cheers Si - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wednesday, September 11, 2002, at 09:59 PM, Simon Blake wrote:
On Wed, Sep 11, 2002 at 09:16:38PM -0400, Joe Abley said:
I ordered the ASNs for the APE and WIX route servers, back in the day.
Not strictly - I obtained the WIX ASN by mine own fair hand, it was the APE one you obtained (and for which we are eternally grateful, since it doesn't cost us anything :-).
Ah, you're right. Sorry :) - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
The reason I mentioned this ... non-RFC compliant idea in the first place, was the number of comments I get from customers who state that the cost and ease of obtaining the AS number is not as Joe says "trivial". It costs a fair bit of money and takes a fair bit of time. I also wanted to say that even though NZ isn't going to fix the global AS number crisis, NZ is not excused from thinking about opportunities to help, if some minor bending of 1930 allows us to do that, then we can at least consider it. The hippocritical part of my proposal is actually the address space requirements, in which I see no easy option other than the allocation of /24's as suggested by Joe. These are easier to come by, and justify as it is often done through the ISP, removing that complexity from the end-user. Thanks for everyone's input, the question I am left with is ... if customer X designing their network with our Routers asks me how to multihome to 2 ISPs, do I say go and talk to APNIC first or forget it, or do I say, the ISPs have jointly agreed on a method of using Private AS numbers, go and talk to them about how it's done ? If we just want to make it happen I would be happy to offer a suitable forum to thrash it out (the three critical components being Beer/Pizza/Chairs). If not I rest my case ... thanks again Arron *********************************************************************** Arron Scott (CCIE #4099) Phone: +64-9-3551951 Systems Engineer Mobile: +64-27-4883163 Cisco New Zealand mailto:ascott(a)cisco.com http://www.cisco.com *********************************************************************** -----Original Message----- From: owner-nznog(a)list.waikato.ac.nz [mailto:owner-nznog(a)list.waikato.ac.nz]On Behalf Of Joe Abley Sent: Thursday, 12 September 2002 10:48 a.m. To: Simon Blake Cc: nznog(a)list.waikato.ac.nz Subject: Re: BGP Private ASes On Thu, Sep 12, 2002 at 10:02:10AM +1200, Simon Blake wrote:
On Wed, Sep 11, 2002 at 05:10:39PM -0400, Joe Abley said:
Just out of interest, what's the benefit in running yet another RPSL database?
WIX has 50+ private ASN peers, AFAIK you can't put info about private ASN into the public RPSL services, so if you want the value RPSL provides for private ASN (and I ohh so do), you run your own database.
... or you use globally-unique ASNs. Using private ASNs for non-private applications is surely broken, especially when it's so trivial to obtain a globally-unique one. You might argue that it's inefficient use of a finite resource for enterprises (in the AS1918 sense) that are not transit providers to be allocated globally-unique ASNs, and I might agree with you. That's not a problem that's going to be solved in New Zealand, though; that's a problem that will be managed by IANA and the RIRs with allocation policies until someone comes up with a multi-homing and routing system that scales better than the one we have.
efficient local mesh at WIX and APE. Perhaps I should consider moving to a provider who...."
... is willing to surrender control of her routing policy to a best-effort coordination service with no responsibility for the quality of the routing data sent to or from her network?
Oh for crying out loud.
I'm not talking about ingress filtering done by the WIX route servers. By "responsibility for the quality", I mean: + having a route propagation path which is different to the packet forwarding path, which is a general problem of route servers on non-trivial layer-2 exchange fabrics; + having no contract/support relationship/whatever between operators connected to the route server, which is a general problem of multi-lateral peering. As to the "best-effort" bit, I thought you were; sorry if that's not the case.
I've pointed this out several times on NZNOG, and yet you continue to assert that it's not the case.
Nope. I don't remember making any comments ever about what ingress policy you were using for the route servers. I do remember making comments about it being generally hard for *other* people to come up with a sensible ingress policy for their session to the route server, though, which is quite different. Lots of people find your route servers useful, and that's great. They *are* useful. I was objecting to the idea that operators who don't use the route servers must be bad or stupid, or be otherwise unworthy of attracting customers, because I don't think that's reasonable; there are arguments for not using them, just as there are arguments *for* using them. Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Wednesday, September 11, 2002, at 09:04 PM, Arron Scott wrote:
The reason I mentioned this ... non-RFC compliant idea in the first place, was the number of comments I get from customers who state that the cost and ease of obtaining the AS number is not as Joe says "trivial". It costs a fair bit of money and takes a fair bit of time.
I ordered an ASN for a project I'm working on, just the other day. I asked my transit provider in NZ, who is an APNIC member, to put the application in to APNIC. It took two e-mails, four working days, and zero dollars.
Thanks for everyone's input, the question I am left with is ... if customer X designing their network with our Routers asks me how to multihome to 2 ISPs, do I say go and talk to APNIC first or forget it, or do I say, the ISPs have jointly agreed on a method of using Private AS numbers, go and talk to them about how it's done ?
The "RFC1930-bending" is a solution looking for a problem. Joe - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
At 13:04 12/09/2002 +1200, Arron Scott wrote:
The reason I mentioned this ... non-RFC compliant idea in the first place, was the number of comments I get from customers who state that the cost and ease of obtaining the AS number is not as Joe says "trivial". It costs a fair bit of money and takes a fair bit of time.
Hmmm, throughout Asia, any organisation who has complained about it being difficult to get any resources from APNIC usually has not tried. Just want to reiterate Joe's experience - it really isn't that hard, and that's the feedback I get when people who said it was hard actually tried. There are two ways of getting an ASN from APNIC: from my multihoming tutorial at the last APNIC conference (www.apnic.net/meetings/14/programme/docs/bgp-tut-slides-pfs.pdf) these are: - existing APNIC member - fill up http://ftp.apnic.net/apnic/docs/asn-request - costs nothing more than you already pay - become an APNIC member (takes some time, and money), then do the above - apply as a non-member - http://www.apnic.net/member/non-member-application.html - cost US$500, annual US$50. Using a private ASN for multihoming of a customer between two ISPs works just fine so long as the two ISPs actually agree that the ASN can be used for that. In my past life I've done that, it's worked. But it definitely falls into the category of "unnecessary" - that's what public ASNs are for.
I also wanted to say that even though NZ isn't going to fix the global AS number crisis, NZ is not excused from thinking about opportunities to help, if some minor bending of 1930 allows us to do that, then we can at least consider it.
What global ASN crisis? http://www.ietf.org/internet-drafts/draft-ietf-idr-as4bytes-05.txt fixes that one, hopefully. :-)
Thanks for everyone's input, the question I am left with is ... if customer X designing their network with our Routers asks me how to multihome to 2 ISPs, do I say go and talk to APNIC first or forget it, or do I say, the ISPs have jointly agreed on a method of using Private AS numbers, go and talk to them about how it's done ? If we just want to make it happen I would be happy to offer a suitable forum to thrash it out (the three critical components being Beer/Pizza/Chairs). If not I rest my case ... thanks again
Go to upstreams - if either is an APNIC member, they get the ASN for them. As per Joe's example, it takes little time, costs nothing, or close to nothing (if the upstream chooses to charge). If upstreams refuse, or are not APNIC/ARIN/RIPE NCC members, the org has the two choices I highlighted above. If they still find they are having problems, let me know, and I will be more than happy to follow up. hth, philip -- - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
participants (8)
-
Andy Linton
-
Arron Scott
-
Ewen McNeill
-
James Tyson
-
Joe Abley
-
Lin Nah
-
Philip Smith
-
Simon Blake