v6 makes things better, because it uses 128-bit addresses rather than 32-bit addresses. See RFCs 1715 and 3194 for the details.
Yes, there's a small linear factor of extra memory required for v6 routes vs v4 routes, but that's irrelevant compared to the route count reduction that comes from a lower HD ratio.
Unless IPv6 addresses are being handed out in a way that's much more conducive to this, it won't really change anything
Which they are, as a direct result of v6 being so huge. See RFCs 1715 and 3194 for discussion on this.
Obviously in the long run we'll end up with a higher absolute count of routes in v6 (because supporting more people was the other reason for it) but the route count will scale far better than a network that has to be run at a ridiculously high HD-ratio because it's too small.
Yeah, it's technically possible in v4, but I need a/27 for my network and I guess most people need/28-/29. Your/19 is only 512/28s, so I guess you aren't going to be giving those out. Realistically, your users are going to end up with 1 v4 IP each, and be stuck with NAT.
(Until you get more than a few thousand customers and have to start CGNATing, at which point they're screwed, especially without v6 as an alternative.)
Who running dual stack gives out blocks of v6 to end users as part of the "standard" residential low-cost service? What sizes are the blocks?
Pretty much everybody does. My ISP gives a/48, and even Comcast give up to/60. (Then there's the depressingly large amount of ISPs that only do one/64, despite RFC 6177 basically saying that you should be getting at least/56...)
I believe dual stack uses resources that increases cost and complexity for the end users, and I don't want to subject them to an inferior service.
Well, it doesn't. If anything the lack of NAT means it uses fewer resources, but nobody will notice anyway because the resource usage of an IP protocol is irrelevantly tiny.
if I have to go out and buy a bunch of v4 addresses for them anyway, why shouldn't I just give them standard v4 Internet?
You know where this logic is headed, right? It won't be long before you won't be able to buy a bunch of v4 addresses for them. Or maybe you will, but they'll be expensive enough to seriously impact your bottom line. Will an extra $5/mo on each customer's bill be enough to count as an inferior experience for them? What about $10/mo? $20/mo? (Not just your bill, mind. Their Netflix bill too, or anybody else they pay for any services that need a server to run. Or maybe the service is free, but shut down because servers are too expensive due to the IP cost.)
That's the future you're trying to get for your users. I don't think it's superior.
If you can't [I'll assume "can"] name a single one that isn't "religious", then I'll reconsider
If you have v6, you can accept inbound connections on any of your computers without dealing with port forwarding/NAT.
If you read my above posts, you'll see I'm not asking you to punish your customers. I'm specifically asking you to do dual stack, so your customers have v6 and Skype works. There is no need to break Skype to get v6 to your customers.
I blame you, and people like you, for refusing to roll out v6. If you'd all just do it, we'd be done by now. Please do your part.
Yes, the situation with Skype sucks, but it's not preventing you from rolling out v6, it's only preventing you from not rolling out v4. We need v6 now; sunsetting v4 has to come after that, not before. Yes, I know it sucks that you can't do v6-only networks yet, but rolling out v6 is step 1 in getting to that point, so please do it.
That's BS. You need v6 to reach other people's v6 servers. Other people need you to have v6 so they can run v6 servers. Your users need v6 so they can run servers that other people can actually connect to without fucking around with NAT. You need v6 for when you inevitably run out of v4 addresses and start having to do CGNAT -- over 50% of your traffic will be on v6, so your CGNAT boxes will only need to handle half the traffic they otherwise would, which makes them cheaper (not to mention your customers will actually be able to receive connections). The internet needs everybody to have v6 because v4 is a clusterfuck with this many users and it's only going to get worse.
You don't need to handle "4-6 NAT" and "6-4 CGNAT" for this. You're overcomplicating it. Just do dual stack. It's easy and it works perfectly fine (and it doesn't break Skype, or anything else).
Skype works just as well on a dual-stack network as it does on a v4 only one. It is broken on v6-only networks even with NAT64 in the picture, which definitely sucks, but please don't let it stop you rolling out v6! Skype is only a blocker if you're trying to remove your v4, which is a separate step that you don't need to be doing yet.
(It's broken because it exchanges v4 literal addresses in the protocol -- there's no space for v6 addresses and it doesn't use DNS, so NAT64+DNS64 is out. Of course MS could fix it easily enough if they could just be bothered to...)
There's also DS-Lite if you really desperately want to run a v6-only access network, but generally dual stack is the way to go, particularly for any network that already exists.
Skype will continue to work. It doesn't care what addresses are used for the management interface, and it doesn't care that you have a dual-stack network -- it just ignores the v6 side. Skype is only a problem if you remove the v4, but you don't need to remove your v4 to deploy v6.
(Or it can be made to work with 464XLAT, if you really want to run a v6-only access network.)
I was assuming that the 32-bit plugin process could be a lightweight shim. I doubt it'd need 64-bit versions of all 54 MB of the dlls that Firefox ships with.
By "full", I mean that it can do DNS. Windows not supporting RDNSS doesn't mean that you have to set the server manually; you can set it automatically.
I'm not really a fan of RDNSS; it puts host config into RAs with no clear guidelines as to which config options ought to be in them. (Why do we only put DNS info in there, and not all the other things you can configure?) But I'm not arguing that MS shouldn't support it, I'm just pointing out that Windows isn't so incapable that it has no way of setting DNS servers automatically.
(As an aside, Windows will also configure a default set of DNS servers if you have no other v6 servers configured, so if you're doing a v6-only network and you really don't want to run stateless DHCPv6 for some reason and the only thing you wanted to set was the DNS servers, you could just add fec0:0:0:ffff::{1,2,3} to your DNS server and Windows would work fine.)
Plus no difference at all between listening to everything you could in the 40s (comparatively not much) and recording what's useful, vs recording everything.
You also said they can't transition to v6 because their own OS doesn't support it, which isn't true. It's supported full automatic configuration of v6 network details out of the box since Vista in 2006, which is a lot longer than most Linux distros have been doing it. I believe Debian only started doing that last year, and I'd be unsurprised if there were still major distros that didn't.
I wish I could find the discussions they must have had at the time about RAs... I assumed there would be mailing list archives or somesuch but I haven't managed to find anything. I guess the logic was that DNS info (or other host config) doesn't belong in RAs, because RAs are broadcasts sent by routers (plural, potentially) to announce network layout. That doesn't match up with the requirements for host config parameters, where you need a single authoritative source and you need the ability to receive machine IDs from clients so you can give out per-machine config settings.
(Of course we haven't really stuck with that logic, since people argued that they didn't want to run DHCPv6 just for DNS, so DNS info was added to RAs. Then other people argued they didn't want to run DHCPv6 just for DNS search domains, so that was added too. Where does it stop, I wonder...)
Um, no. We're running out of addresses because we don't have enough addresses.
(And to address some of the other misunderstandings: ARIN still have a v6 printing press, v6 doesn't magically expose everything to the entire WWW, you can still run a central firewall in v6 (just without NAT, thankfully) and your IPs won't require memorizing 128 bits unless you're dumb enough to pick an address that uses all 128 bits, in which case you don't get to complain about it).
v6 doesn't require you to memorize 8 groups of 4 letters. You can put v6 addresses into/etc/hosts to avoid having to remember them... and if you find that syncing a huge/etc/hosts file around is a pain, then it also supports that newfangled "DNS" thing to save you the effort.
Plus you can set any bits in your allocation to zero, and your allocation should be at least/56 or so, which means the number of bits you actually have to remember is about the same as in the v4 case (where you have to remember a 32-bit RC1918 IP plus the 32-bit global IP), so it's not really any worse than the current situation.
That's an odd definition of "already", given that it came years and years after v4 was extended to 128 bits. Also the 128-bit version is actually big enough to handle the number of hosts we need it to handle, and it has far wider support and deployment than that 64-bit extension.
Because it's now being aimed very explicitly at people too dumb to figure out how to use a browser, and they're deliberately dropping anything that requires any thought to figure out. That was not the case previously. Is it such an unreasonable stretch of the imagination that somebody that picked a browser that catered for users with a brain cell might be unhappy at being moved to a browser that alienates such users?
(Also there's an ongoing and very serious effort to trash system themes and go their own way. If you decided to install Camino, would you have been very happy at being silently moved to upstream Firefox?)
But hey, sure, the back button's a circle. Totally the problem.
Ah, that's unfortunate. It's been causing lots of random crashes in GC-related functions for some people, so disabling it could've been an easy fix. And as you say, it's unlikely your hardware went bad at exactly the time you upgraded (although it's possible for new versions of software to reveal existing problems).
The only other great idea I have, if you haven't tried it already, is to disable hardware acceleration. It's possible your previous drivers were blacklisted (disabling acceleration) and the new ones aren't, but ought to be for whatever reason (flaky GPU hardware being one possibility).
Slashdot Mirror
v6 makes things better, because it uses 128-bit addresses rather than 32-bit addresses. See RFCs 1715 and 3194 for the details.
Yes, there's a small linear factor of extra memory required for v6 routes vs v4 routes, but that's irrelevant compared to the route count reduction that comes from a lower HD ratio.
Unless IPv6 addresses are being handed out in a way that's much more conducive to this, it won't really change anything
Which they are, as a direct result of v6 being so huge. See RFCs 1715 and 3194 for discussion on this.
Obviously in the long run we'll end up with a higher absolute count of routes in v6 (because supporting more people was the other reason for it) but the route count will scale far better than a network that has to be run at a ridiculously high HD-ratio because it's too small.
Yeah, it's technically possible in v4, but I need a /27 for my network and I guess most people need /28-/29. Your /19 is only 512 /28s, so I guess you aren't going to be giving those out. Realistically, your users are going to end up with 1 v4 IP each, and be stuck with NAT.
(Until you get more than a few thousand customers and have to start CGNATing, at which point they're screwed, especially without v6 as an alternative.)
Who running dual stack gives out blocks of v6 to end users as part of the "standard" residential low-cost service? What sizes are the blocks?
Pretty much everybody does. My ISP gives a /48, and even Comcast give up to /60. (Then there's the depressingly large amount of ISPs that only do one /64, despite RFC 6177 basically saying that you should be getting at least /56...)
I believe dual stack uses resources that increases cost and complexity for the end users, and I don't want to subject them to an inferior service.
Well, it doesn't. If anything the lack of NAT means it uses fewer resources, but nobody will notice anyway because the resource usage of an IP protocol is irrelevantly tiny.
if I have to go out and buy a bunch of v4 addresses for them anyway, why shouldn't I just give them standard v4 Internet?
You know where this logic is headed, right? It won't be long before you won't be able to buy a bunch of v4 addresses for them. Or maybe you will, but they'll be expensive enough to seriously impact your bottom line. Will an extra $5/mo on each customer's bill be enough to count as an inferior experience for them? What about $10/mo? $20/mo? (Not just your bill, mind. Their Netflix bill too, or anybody else they pay for any services that need a server to run. Or maybe the service is free, but shut down because servers are too expensive due to the IP cost.)
That's the future you're trying to get for your users. I don't think it's superior.
If you can't [I'll assume "can"] name a single one that isn't "religious", then I'll reconsider
If you have v6, you can accept inbound connections on any of your computers without dealing with port forwarding/NAT.
There are others, but whatever, there's one.
If you read my above posts, you'll see I'm not asking you to punish your customers. I'm specifically asking you to do dual stack, so your customers have v6 and Skype works. There is no need to break Skype to get v6 to your customers.
I blame you, and people like you, for refusing to roll out v6. If you'd all just do it, we'd be done by now. Please do your part.
Yes, the situation with Skype sucks, but it's not preventing you from rolling out v6, it's only preventing you from not rolling out v4. We need v6 now; sunsetting v4 has to come after that, not before. Yes, I know it sucks that you can't do v6-only networks yet, but rolling out v6 is step 1 in getting to that point, so please do it.
That's BS. You need v6 to reach other people's v6 servers. Other people need you to have v6 so they can run v6 servers. Your users need v6 so they can run servers that other people can actually connect to without fucking around with NAT. You need v6 for when you inevitably run out of v4 addresses and start having to do CGNAT -- over 50% of your traffic will be on v6, so your CGNAT boxes will only need to handle half the traffic they otherwise would, which makes them cheaper (not to mention your customers will actually be able to receive connections). The internet needs everybody to have v6 because v4 is a clusterfuck with this many users and it's only going to get worse.
You don't need to handle "4-6 NAT" and "6-4 CGNAT" for this. You're overcomplicating it. Just do dual stack. It's easy and it works perfectly fine (and it doesn't break Skype, or anything else).
Skype works just as well on a dual-stack network as it does on a v4 only one. It is broken on v6-only networks even with NAT64 in the picture, which definitely sucks, but please don't let it stop you rolling out v6! Skype is only a blocker if you're trying to remove your v4, which is a separate step that you don't need to be doing yet.
(It's broken because it exchanges v4 literal addresses in the protocol -- there's no space for v6 addresses and it doesn't use DNS, so NAT64+DNS64 is out. Of course MS could fix it easily enough if they could just be bothered to...)
There's also DS-Lite if you really desperately want to run a v6-only access network, but generally dual stack is the way to go, particularly for any network that already exists.
Yeah, don't bother with that. Deploy v6 now, move to CGNAT for v4 when you have to (possibly soonish), and then worry about anything else much later.
Skype will continue to work. It doesn't care what addresses are used for the management interface, and it doesn't care that you have a dual-stack network -- it just ignores the v6 side. Skype is only a problem if you remove the v4, but you don't need to remove your v4 to deploy v6.
(Or it can be made to work with 464XLAT, if you really want to run a v6-only access network.)
I was assuming that the 32-bit plugin process could be a lightweight shim. I doubt it'd need 64-bit versions of all 54 MB of the dlls that Firefox ships with.
This sounds like an actual use for that stub installer that you serve by default to Windows users. Just have it pick which version to download.
Its user base would rather help than be snarky, but it's hard to help when Mozilla completely rejects you.
More on-topic, I found this: https://groups.google.com/forum/#!msg/mozilla.dev.platform/oB-GAXt6Ijo/lUgjfUZ8ArEJ. Apparently "working on it" is the new way to spell "ignoring".
By "full", I mean that it can do DNS. Windows not supporting RDNSS doesn't mean that you have to set the server manually; you can set it automatically.
I'm not really a fan of RDNSS; it puts host config into RAs with no clear guidelines as to which config options ought to be in them. (Why do we only put DNS info in there, and not all the other things you can configure?) But I'm not arguing that MS shouldn't support it, I'm just pointing out that Windows isn't so incapable that it has no way of setting DNS servers automatically.
(As an aside, Windows will also configure a default set of DNS servers if you have no other v6 servers configured, so if you're doing a v6-only network and you really don't want to run stateless DHCPv6 for some reason and the only thing you wanted to set was the DNS servers, you could just add fec0:0:0:ffff::{1,2,3} to your DNS server and Windows would work fine.)
Plus no difference at all between listening to everything you could in the 40s (comparatively not much) and recording what's useful, vs recording everything.
You also said they can't transition to v6 because their own OS doesn't support it, which isn't true. It's supported full automatic configuration of v6 network details out of the box since Vista in 2006, which is a lot longer than most Linux distros have been doing it. I believe Debian only started doing that last year, and I'd be unsurprised if there were still major distros that didn't.
I wish I could find the discussions they must have had at the time about RAs... I assumed there would be mailing list archives or somesuch but I haven't managed to find anything. I guess the logic was that DNS info (or other host config) doesn't belong in RAs, because RAs are broadcasts sent by routers (plural, potentially) to announce network layout. That doesn't match up with the requirements for host config parameters, where you need a single authoritative source and you need the ability to receive machine IDs from clients so you can give out per-machine config settings.
(Of course we haven't really stuck with that logic, since people argued that they didn't want to run DHCPv6 just for DNS, so DNS info was added to RAs. Then other people argued they didn't want to run DHCPv6 just for DNS search domains, so that was added too. Where does it stop, I wonder...)
It's not that hard, but it's not that easy either. It's far simpler without NAT, where you just connect to the machine.
Also it suddenly gets very, very hard when your ISP puts you behind their own NAT, so you don't even have a "public IP" for your laptop to connect to.
Um, no. We're running out of addresses because we don't have enough addresses.
(And to address some of the other misunderstandings: ARIN still have a v6 printing press, v6 doesn't magically expose everything to the entire WWW, you can still run a central firewall in v6 (just without NAT, thankfully) and your IPs won't require memorizing 128 bits unless you're dumb enough to pick an address that uses all 128 bits, in which case you don't get to complain about it).
v6 doesn't require you to memorize 8 groups of 4 letters. You can put v6 addresses into /etc/hosts to avoid having to remember them... and if you find that syncing a huge /etc/hosts file around is a pain, then it also supports that newfangled "DNS" thing to save you the effort.
Plus you can set any bits in your allocation to zero, and your allocation should be at least /56 or so, which means the number of bits you actually have to remember is about the same as in the v4 case (where you have to remember a 32-bit RC1918 IP plus the 32-bit global IP), so it's not really any worse than the current situation.
That's an odd definition of "already", given that it came years and years after v4 was extended to 128 bits. Also the 128-bit version is actually big enough to handle the number of hosts we need it to handle, and it has far wider support and deployment than that 64-bit extension.
So really, there's no point in it.
No we won't. Anybody who thinks this doesn't understand how large 2^128 is.
(If you disagree with me, try to back it up with actual numbers.)
Except it can, because it supports stateless DHCPv6 (unless you're on XP).
Because it's now being aimed very explicitly at people too dumb to figure out how to use a browser, and they're deliberately dropping anything that requires any thought to figure out. That was not the case previously. Is it such an unreasonable stretch of the imagination that somebody that picked a browser that catered for users with a brain cell might be unhappy at being moved to a browser that alienates such users?
(Also there's an ongoing and very serious effort to trash system themes and go their own way. If you decided to install Camino, would you have been very happy at being silently moved to upstream Firefox?)
But hey, sure, the back button's a circle. Totally the problem.
Ah, that's unfortunate. It's been causing lots of random crashes in GC-related functions for some people, so disabling it could've been an easy fix. And as you say, it's unlikely your hardware went bad at exactly the time you upgraded (although it's possible for new versions of software to reveal existing problems).
The only other great idea I have, if you haven't tried it already, is to disable hardware acceleration. It's possible your previous drivers were blacklisted (disabling acceleration) and the new ones aren't, but ought to be for whatever reason (flaky GPU hardware being one possibility).
Who will want to use a browser that the loyal fans seem to hate?
Well, exactly. And whose fault is it that loyal fans hate it?