Re: [nznog] THE SKY IS FALLING ( was Re: IPv4 Exhaustion)
: The past 10+ years of work on IPv6 nee IPng have taught : many lessons about what will and will not work (and more : importantly what will be accepted or not). Exactly. I NEED to multihome. I won't accept not being able to do that. No provider is going to be good enough that I trust ALL of my connectivity with them and no one else. That settles it right there. A protocol should not dictate my business practice. : ps - the 100-million number likely came from Alain Durand's : presentation "Managing 100+ Million IP Addresses" WRT "Managing 100+ Million IP Addresses", I see the following chart. (apologies for the formatting. It's on page 6 at www.nanog.org/mtg-0606/pdf/alain-durand.pdf
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Triple Play Effect on the Use of IP Addresses 2005 2006+ HSD only Triple Play
Cable Modem 1(private only) 1 Home Computer/Router 1 1 eMTA (Voice adaptor) 0 1 – 2 Set Top Box (STB) 0 2 Total number of IP addresses(assume 2.5 1 – 2 8 – 9 STB per household)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
What STB needs 2 IPs per box? The way we're doing it is one private IP per STB and one STB per television set, so that's 2-3 private IPs per household. One public IP per customer for Internet and we're not going for VoIP until later. Perhaps they have one STB per household no matter the number of television sets? Also, I see they have 20-30 ASNs. Are all 20M customers going to be in one AS? I wouldn't expect that.
Finally, with VPRNs addresses could be reused. Just put the customers in different VPRNs and duplicate the address scheme. It's 'divide and conquer' methodology. I'm a noobie to VPRNs, so flame me if this isn't feasible. I have doused my flameproof underpants with beer, so I'm protected... :-)
scott
--- jejs+lists(a)sahala.org wrote:
From: "joshua sahala"
Hi, Good point on the number of addresses required for STBs. In a common and well deployed approach to deploying multiplay solutions (ie. video and other stuff), a single public address is required at the residential gateway. It is perfectly sane to use 0.0.0.0 or some other unique address to receive the multicast traffic (video). A residential gateway that is performing IGMP proxy / snooping will have the capability to send it's IGMP joins out the IPoE or IPoA interface and it does not need a valid routable address for this traffic. Since the RG is performing IGMP proxy, it is possible to deploy multiple STBs within a household that all have private addresses. best regards, truman On 29/11/2006, at 12:07 PM, Scott Weeks wrote:
: The past 10+ years of work on IPv6 nee IPng have taught : many lessons about what will and will not work (and more : importantly what will be accepted or not).
Exactly. I NEED to multihome. I won't accept not being able to do that. No provider is going to be good enough that I trust ALL of my connectivity with them and no one else. That settles it right there. A protocol should not dictate my business practice.
: ps - the 100-million number likely came from Alain Durand's : presentation "Managing 100+ Million IP Addresses"
WRT "Managing 100+ Million IP Addresses", I see the following chart. (apologies for the formatting. It's on page 6 at www.nanog.org/mtg-0606/pdf/alain-durand.pdf
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Triple Play Effect on the Use of IP Addresses 2005 2006+ HSD only Triple Play
Cable Modem 1(private only) 1 Home Computer/Router 1 1 eMTA (Voice adaptor) 0 1 – 2 Set Top Box (STB) 0 2 Total number of IP addresses(assume 2.5 1 – 2 8 – 9 STB per household)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
What STB needs 2 IPs per box? The way we're doing it is one private IP per STB and one STB per television set, so that's 2-3 private IPs per household. One public IP per customer for Internet and we're not going for VoIP until later. Perhaps they have one STB per household no matter the number of television sets? Also, I see they have 20-30 ASNs. Are all 20M customers going to be in one AS? I wouldn't expect that.
Finally, with VPRNs addresses could be reused. Just put the customers in different VPRNs and duplicate the address scheme. It's 'divide and conquer' methodology. I'm a noobie to VPRNs, so flame me if this isn't feasible. I have doused my flameproof underpants with beer, so I'm protected... :-)
scott
On Wed, 29 Nov 2006, Truman Boyes wrote:
Good point on the number of addresses required for STBs. In a common and well deployed approach to deploying multiplay solutions (ie. video and other stuff), a single public address is required at the residential gateway. It is perfectly sane to use 0.0.0.0 or some other unique address to receive the multicast traffic (video).
No, don't use 0.0.0.0. Because some devices when configured this way will reply to ARP requests for any address. Which makes for a large amount of background ARP noise. --David
Scott! On 28-Nov-2006, at 18:07, Scott Weeks wrote:
: The past 10+ years of work on IPv6 nee IPng have taught : many lessons about what will and will not work (and more : importantly what will be accepted or not).
Exactly. I NEED to multihome. I won't accept not being able to do that. No provider is going to be good enough that I trust ALL of my connectivity with them and no one else. That settles it right there. A protocol should not dictate my business practice.
You can multi-home in v6 in precisely the same way as you do in v4 if you're an ISP, under all RIR policies. This also goes for anybody who qualified for v4 PI assignments in the ARIN region (so, end users too, at least those who qualify for sufficient addresses to want to multi-home with PI v4 space). Other regions will presumably follow suit if their respective memberships want that to happen. The ability to multi-home using PA v6 space (as is commonly done in the v4 network) depends on deployed filtering practice. Currently, it doesn't work very well; however, if there was pressure from an actual customer base, it could well be that it would work more often. Neither of these require any additional protocol development or implementation effort, and both ought to be very familiar to anybody who is multi-homed using v4. Nothing until this sentence has had anything to do with shim6, for example. There are lots of arguments for why v6 won't/shouldn't/can't succeed as a replacement for v4, but multi-homing is barely one of them. Joe
Scott Weeks wrote:
Finally, with VPRNs addresses could be reused. Just put the customers in different VPRNs and duplicate the address scheme. It's 'divide and conquer' methodology. I'm a noobie to VPRNs, so flame me if this isn't feasible. I have doused my flameproof underpants with beer, so I'm protected... :-)
Oh yes. This is a valid methodology. Also, you don't really need every subscriber to be uniquely addressed across your entire network - to the closest PoP site or Super-PoP site is sufficient, depending on your topology. Of course, from your Ethernet switch/DSLAM/FTTx node/whatever, you know where and who the subscriber is, so you're not reliant on IP addressing to authorize content access, either.... aj. (just scrolling back through the messages, thought I'd add that since it was an actual question and nobody answered it.) -- nobody's still really given me a good reason about what problems v6 is solving. hammer, nail, etc.
I don't know how more obvious it can be made. It's solving the looming problem of the IPv4 address pool being exhausted in the next 3 to 6 years. Is not complete lack of Internet connectivity for new customers not a good enough reason? -----Original Message----- nobody's still really given me a good reason about what problems v6 is solving. hammer, nail, etc.
Philip D'Ath wrote:
I don't know how more obvious it can be made.
It's solving the looming problem of the IPv4 address pool being exhausted in the next 3 to 6 years.
Is not complete lack of Internet connectivity for new customers not a good enough reason?
I think the subject line says it all, on that particular topic. V6 seems to largely be a problem looking for a solution at this point in time. Need to support 25 million CPE devices? There are ways to do that, and many vendors are falling over themselves to help you. STBs? No problem. Triple/quad/quinta-plays? Got it covered. While I don't doubt that IPv6 will eventually become the widespread protocol, do you REALLY see that happening before 2020-ish? At which point the carriers will be able to make a premium selling v4 service, because by gosh there will be a lot of legacy networks out there needing it. But that's a completely separate topic. aj.
Philip D'Ath wrote:
It's solving the looming problem of the IPv4 address pool being exhausted in the next 3 to 6 years.
Is not complete lack of Internet connectivity for new customers not a good enough reason?
Sadly v6 is at best only a partial solution. If v6 really allowed a customer to do all the things that a v4 address does then there would be no shortage of money and people deploying it. The only thing that I can think of that would drive widespread v6 deployment would be compelling content or services that are simply not available on v4. If you were a content or service provider would you keep your service v6 only? I think not. Lack of v4 addresses for new comers or those who can get (think buy) them will not drive v6 deployment. More likely it will drive ISP aggregation and address trading whether legal or otherwise. What is really needed here is a migration tool that will allow seamless deployment of v6. That's seamless in the sense that a user does not notice or care which type of address they get. I suspect there's a buck in it for anyone who can work that out. -- Robert Gray bob(a)brockhurst.co.nz
We'll have to disagree on this one. IPv6 offers many improvements over IPv4, apart from increased address space. For example, the whole concept of broadcasts has gone (there are none). Protocols like arp are gone. Many people wont need to use DHCP any more. I've said it before (and it should be pointed out this is my opinion, which differs from others on this matter) that once IPv4 address space is gone there will be NO internet access for new customers. So the compelling driver would be for new people to be able to access the Internet. I don't think you are appreciating the seriousness of this. Either you will Internet access with IPv6, or no access whatsoever - no content, nothing. If people commence using dual stacks today, the migration to IPv6 will be seamless. There is no reason for end users today to notice any change. If it is left for 6 years there will be a massive upheaval. -----Original Message----- ... Sadly v6 is at best only a partial solution. If v6 really allowed a customer to do all the things that a v4 address does then there would be no shortage of money and people deploying it. The only thing that I can think of that would drive widespread v6 deployment would be compelling content or services that are simply not available on v4. If you were a content or service provider would you keep your service v6 only? I think not. Lack of v4 addresses for new comers or those who can get (think buy) them will not drive v6 deployment. More likely it will drive ISP aggregation and address trading whether legal or otherwise. What is really needed here is a migration tool that will allow seamless deployment of v6. That's seamless in the sense that a user does not notice or care which type of address they get. I suspect there's a buck in it for anyone who can work that out.
On Thu, 30 Nov 2006, Robert Gray wrote:
Sadly v6 is at best only a partial solution. If v6 really allowed a customer to do all the things that a v4 address does then there would be no shortage of money and people deploying it.
Excuse my ignorance, but what, exactly, can one do with v4 that can't be done with v6? Other than giving people relatively-easily-remembered addresses, that is :P This is a serious question. I know that there's no such thing as broadcasting with v6, but is that such a terrible loss? What other superiorities does v4 exhibit? -- Matthew Poole "Don't use force. Get a bigger hammer."
Most of the things removed from IPv6 were done to make the system more scalable. For example, there is no broadcast mechanism. So applications that rely on using broadcasts need to be re-written (to use multicast). Applications that rely on scanning IP blocks to detect things will need to be re-written, as each subnet is now impossibly large to scan. -----Original Message----- ... Excuse my ignorance, but what, exactly, can one do with v4 that can't be done with v6? Other than giving people relatively-easily-remembered addresses, that is :P
For example, there is no broadcast mechanism. So applications that rely on using broadcasts need to be re-written (to use multicast). Applications that rely on scanning IP blocks to detect things will need to be re-written, as each subnet is now impossibly large to scan.
channeling herr bush... "I encourage my competition to believe that" --bill
Matthew Poole wrote:
Excuse my ignorance, but what, exactly, can one do with v4 that can't be done with v6? Other than giving people relatively-easily-remembered addresses, that is :P
Maybe I've got this wrong, if so my apologies and I'd be happy to be corrected. I've understood that if I only have a v6 IP address: *. my web site would only be visible to others with v6 addresses (rather defeating the point) *. I would only get mail on my mail server from others with v6 addresses (which I think may limit the amount of mail I get) *. P2P apps would only work with others who have v6 addresses (so no webcam/skype/SIP with the family for Christmas) *. I think I'd have some problems with the DNS as not all servers will respond to v6 queries (I suspect this is minor and could be easily solved) In other words I could not play properly with the v4 internet. Again if I have it right I could establish sessions to v4 addresses in the same way as I can using NAT from my RFC1918 address so web browsing, mail delivery and so forth would be fine. -- Robert Gray bob(a)brockhurst.co.nz
On Thu, 30 Nov 2006, Robert Gray wrote:
Matthew Poole wrote:
Excuse my ignorance, but what, exactly, can one do with v4 that can't be done with v6? Other than giving people relatively-easily-remembered addresses, that is :P
*SNIP*
In other words I could not play properly with the v4 internet.
But that's not anything that v4 can do that v6 can't, inasmuch as it's to do with content availability rather than limitations of v6. All your examples also work perfectly conversely in a v6 world where the odd-one-out is a person with a v4 address - that is, in such a world the v4 person would be SoL for getting much done. I've still not seen a single example of something actually worthwhile that can be done with v4 but not with v6. That applications relying on broadcast addreses in v4 will have to be rewritten is not a "v6 cannot do xyz thing that v4 can" situation, since v6 just does it differently rather than not at all. So, again, what does v4 do that v6 cannot, under any circumstances, do? -- Matthew Poole "Don't use force. Get a bigger hammer."
Matthew Poole wrote:
So, again, what does v4 do that v6 cannot, under any circumstances, do?
Clearly from a technical perspective nothing, problems only arise in a mixed v4/v6 world. Hence my earlier comments about a simultaneous cutover from v4 to v6 being required. My statement that v6 is at best only a partial solution is based on the observation that a progressive deployment is the only deployment mechanism that could work. My contention is that this progressive deployment is not commercially, or indeed functionally, acceptable and that far from paying a premium for v6 addresses some (many?) users would resist (probably quite strongly) any move to migrate. -- Robert Gray bob(a)brockhurst.co.nz
On 29-Nov-2006, at 13:46, Robert Gray wrote:
Philip D'Ath wrote:
It's solving the looming problem of the IPv4 address pool being exhausted in the next 3 to 6 years.
Is not complete lack of Internet connectivity for new customers not a good enough reason?
Sadly v6 is at best only a partial solution. If v6 really allowed a customer to do all the things that a v4 address does then there would be no shortage of money and people deploying it.
v6 really doesn't stop anybody doing anything they can do with v4. As one of Wood's recent batch of t-shirts says, "96 more bits. no magic." The main operational differences derive from policy, not protocols. While the IETF and the RIRs might suggest particular rules for deaggregation, summarisation, filtering, etc at the end of the day it's the network operators who get to decide how they run their own networks. (and it's not like people don't get plenty of chances to influence the work of the IETF and the RIRs, anyway) Joe
participants (9)
-
Alastair Johnson -
bmanning＠karoshi.com -
David Robb -
Joe Abley -
Matthew Poole -
Philip D'Ath -
Robert Gray -
Scott Weeks -
Truman Boyes