Any IPV6 Gurus Here?

I’m writing an article for an industry trade magazine and I’m looking at IPV6. I understand the 128-bit vs. 32-bits; I even understand (at least in concept) how IPV4 addresses can be “translated” to IPV6 values. Most of the articles that I’ve run across (including the Wikipedia entries) devote a mind-numbing amount of time to (over-)explaining this part, which is actually pretty straightforward.

What I don’t understand is where the traditional port numbers come into play.

For example, suppose I have a server that’s running Apache (port 80), sendmail (port 25) and a couple of other services. Right now, I simply use a router to NAT those incoming requests from our IPV4 public address to that server box’s internal IPV4 address. Neat and simple.

But I’ve run across a lot of confusing information about IPV6 on the Internet. Some references say that IPV6 doesn’t HAVE ports; that you will alias a separate IPV6 address for each service on that server (which seems stupid to me!). Others say that IETF has agreed to permit IPV4-style NAT now, after strongly resisting it in the past.

Anyone here know? Anyone have a link to something that explains it in plain English (yes, I know that the RFC’s are the final authority, but surely someone has a simple, “here’s how you convert your server box from IPV4 to IPV6” walkthrough on the Web)?

Thanks in advance!

And here I am, replying to myself again! Truly, I live a schizophrenic existence. :slight_smile:

I had forgotten about the difference between the “IP” and the “TCP” part of the packet. The IP header doesn’t contain that info, and the port number is in the TCP header.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I’m pretty sure that’s wrong… IPv4 and IPv6 is all layer three stuff,
and ports are all TCP (layer 4) so one does not necessarily mean the end
of the other. The references I’ve read still include ports after IPv6
addresses.

Good luck.

On 11/04/2010 03:36 PM, smpoole7 wrote:
>
> I’m writing an article for an industry trade magazine and I’m looking at
> IPV6. I understand the 128-bit vs. 32-bits; I even understand (at least
> in concept) how IPV4 addresses can be “translated” to IPV6 values. Most
> of the articles that I’ve run across (including the Wikipedia entries)
> devote a mind-numbing amount of time to (over-)explaining this part,
> which is actually pretty straightforward.
>
> What I don’t understand is where the traditional port numbers come into
> play.
>
> For example, suppose I have a server that’s running Apache (port 80),
> sendmail (port 25) and a couple of other services. Right now, I simply
> use a router to NAT those incoming requests from our IPV4 public address
> to that server box’s internal IPV4 address. Neat and simple.
>
> But I’ve run across a lot of confusing information about IPV6 on the
> Internet. Some references say that IPV6 doesn’t HAVE ports; that you
> will alias a separate IPV6 address for each service on that server
> (which seems stupid to me!). Others say that IETF has agreed to permit
> IPV4-style NAT now, after strongly resisting it in the past.
>
> Anyone here know? Anyone have a link to something that explains it in
> plain English (yes, I know that the RFC’s are the final authority, but
> surely someone has a simple, “here’s how you convert your server box
> from IPV4 to IPV6” walkthrough on the Web)?
>
> Thanks in advance!
>
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.15 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=w7xd
-----END PGP SIGNATURE-----

And the UDP header. The stuff about ports will remain the same, ignore any articles that say ports are going away.

Some of the uses of NAT (need to conserve IP addresses) will be unnecessary in IPv6, but NAT is still useful in other places.

I agree with the others. The ports wont go away. You can see more information on this by reading a really dry article. RFC 2460 http://www.ietf.org/rfc/rfc2460.txt This may also help; Netfilter Extensions HOWTO: New IPv6 netfilter matches or even this ip6tables: IPv6 Firewall For Linux

smpoole7 wrote:

>
> I’m writing an article for an industry trade magazine and I’m looking at
> IPV6. I understand the 128-bit vs. 32-bits; I even understand (at least
> in concept) how IPV4 addresses can be “translated” to IPV6 values. Most
> of the articles that I’ve run across (including the Wikipedia entries)
> devote a mind-numbing amount of time to (over-)explaining this part,
> which is actually pretty straightforward.
>
> What I don’t understand is where the traditional port numbers come into
> play.
>
> For example, suppose I have a server that’s running Apache (port 80),
> sendmail (port 25) and a couple of other services. Right now, I simply
> use a router to NAT those incoming requests from our IPV4 public address
> to that server box’s internal IPV4 address. Neat and simple.
>
> But I’ve run across a lot of confusing information about IPV6 on the
> Internet. Some references say that IPV6 doesn’t HAVE ports; that you
> will alias a separate IPV6 address for each service on that server
> (which seems stupid to me!). Others say that IETF has agreed to permit
> IPV4-style NAT now, after strongly resisting it in the past.
>
> Anyone here know? Anyone have a link to something that explains it in
> plain English (yes, I know that the RFC’s are the final authority, but
> surely someone has a simple, “here’s how you convert your server box
> from IPV4 to IPV6” walkthrough on the Web)?
>

The ideal… you move to IPv6. Your addresses, like everyone’s is
routable… there is NO NAT… goes back to the way it was way back when.

You’ll use your network to control which of you fully routable hosts are
actually visible to the Internet.

You WILL have to replace all of your network equipment (likely… unless
you’ve already been planning for IPv6 with regards to your network
purchases).

Thanks for the replies, folks.

CJ,

I chatted (via email) with my ISP’s tech gurus yesterday and got a better perspective on this. IPV6 is still many, many years away if you’re talking about LAN, especially small networks. IPV6 is already being used on the big backbone links between cities, but is translated to IPV4 at the local level. Another approach, for those ISPs who are still set up for IPV4, is that the IPV4 is being “tunneled” over the IPV6 backbone links.

My ISP said the big holdup in the states isn’t really the equipment; they’re not providing IPV6 service to their customers yet, but they have great big Juniper routers that are IPV6-ready. (Cool looking, too. I saw them when we co-located our mail server in their Geek Room a couple of months ago). :slight_smile:

Another thing that they will have to hammer out, and which is still being worked on, is DNS. Especially at first, there will have to be two DNS systems, one for IPV4 and one for IPV6 (with a LOT of translation between the two – for example, you’re IPV6, but you’re looking up a domain that’s on an IPV4-only subnet).

The one thing that they assured me, and that I’ll pass along here, is that the basics are standardized and agreed upon, but there are still many fine details that aren’t etched in stone yet. The thing about the ports that confused me, for example, came from one reference that I read online. It stated emphatically that ports wouldn’t be used with IPV6 (and yes, it’s completely inaccurate). Another reference said that stateful filtering and NAT weren’t available yet for IPV6; that’s not true, either, because that capability has been present since Linux kernel 2.6.0 (IIRC).

We are headed toward an interesting future. :slight_smile:

Oh, and moderators, if this thread should be moved over to Chit-Chat or Soapbox, please feel free to do so, with my apologies.

The cursory info I’ve heard is that <all> equipment and OS manufactured and sold for about the past 5 years are IPv6 “ready” (that probably means capable, but how easy to enable might be another matter).

From a practical sense, it’s probably safe to observe that the world hasn’t yet been taken over by technology where every man/woman/child will need 6 personal, routable IP addresses which would exhaust available IPv4 addresses, and if that world should arrive IPv6 will probably get a fairly decent “baked in” period beforehand.

BTW - If they will have to re-engineer parts of DNS for Ipv6, I’d all be for junking and re-architecting the whole thing from the ground up, it’s well known likely be “the” Achilles’s heel of the Internet.

IMO,
Tony

There are various problems with using all of the theoretical IPv4 space, which you can read here:

IPv4 address exhaustion - Wikipedia, the free encyclopedia

According to that, less than 5% remains as of today.

IPv4 addresses are unevenly distributed. So it’s understandable that countries like China that missed out on large slabs of IPv4 address space because they were “not around” when the Internet was invented are at the forefront of the move to IPv6.

Also there won’t be a “changeover day”. The technologies will coexist for a long time.

So it’s not as simplistic as 2^32 / (population of world).

tsu2 wrote:

>
> The cursory info I’ve heard is that <all> equipment and OS manufactured
> and sold for about the past 5 years are IPv6 “ready” (that probably
> means capable, but how easy to enable might be another matter).

:slight_smile: Actually, yes and no. You might have a IPv6 capable switch/router, but
once you implement, you might find that you are VERY lacking in memory on
the device (just an example). Remember, at work, we’ve been running IPv6
for years… but most people/companies have a hard time affording the class
of equipment we use. Right now we are using Cisco Nexus so we can go 10Gbit
to the hosts.

The big issue will be the home gateway/routers. I expect the Feds to do a
“voucher” much like what they did for ATSC when they killed VHF/UHF NTSC in
the USA. You’ll just have to trust me… BILLIONS of dollars of equipment
will need to be replaced. And most companies are NOT ready… again, you’ll
just have to trust me on that one. Switching to IPv6 may be the most
expensive (mandated) move in history.

>
> From a practical sense, it’s probably safe to observe that the world
> hasn’t yet been taken over by technology where every man/woman/child
> will need 6 personal, routable IP addresses which would exhaust
> available IPv4 addresses, and if that world should arrive IPv6 will
> probably get a fairly decent “baked in” period beforehand.
>
> BTW - If they will have to re-engineer parts of DNS for Ipv6, I’d all
> be for junking and re-architecting the whole thing from the ground up,
> it’s well known likely be “the” Achilles’s heel of the Internet.

I run our DNS… it speaks IPv6 fluently and handles IPv6 records just fine.
So… not a problem. But as with ALL of IPv6, there’s a price for handling
larger amounts of data. Most all routers/switches are optimized for IPv4…
NOT for IPv6 (even the new ones with IPv6 support).

If every device comes with 100 addresses… and your door knob, your shoe
and the hairs on your head all have addresses… I dare say, we’ll need a
super DNS somehow. I imagine Microsoft will offer to house all IPv6 dbs for
us :slight_smile:

DNS… it speaks IPv6 fluently and handles IPv6 records just fine.
So… not a problem. But as with ALL of IPv6, there’s a price for handling
larger amounts of data. Most all routers/switches are optimized for IPv4…
NOT for IPv6 (even the new ones with IPv6 support).

Capacity and throughput are issues, but what about security?
It’s well known that our current DNS system is subject to a number of unaddressable vulnerabilities, there are several attacks which can only be mitigated once recognized, everything from DoS to cache poisoning. So much of the existing system relies on trusting… Trusting everyone in the DNS chain to properly secure the data, trusting that client machines are who they say they are, trusting that they will behave “properly.”

DNS attacks have taken down some large sites in the past, and AFAIK those same attacks are still threats and perhaps even more so if the target doesn’t have the somewhat specialized ability to defend.

So, I’m all for a complete re-build from the ground up if it’ll fix known security issues and maybe the switchover to IPv6 is the right time (disruption anyway).

Tony

It may not be necessary. Developed countries have larger amounts of IPv4 space left. In less developed countries they can roll out IPv6 as new customers get coverage. Also those home routers have short working lives, a couple of years, because people are changing them to get more features, VoIP, 802.11n, etc.

Another place IPv6 is making inroads is mobile devices.

As for the switchover to digital TV, a friend of mine commented that the US went about it the wrong way, mandating a switchover date. Most other countries brought it in gradually. Here the transition started with an isolated area that was relatively easy to switch and the process won’t be complete until 2013. That way, problems get ironed out along the way, instead of mass panic.

And in fact the IPv6 transition is meant to be gradual takeover.

Well, heck, I’ll ask you, then. How IS the DNS transition going to be done? I’m thinking that eventually, when the available IPV4 addresses run out in the United States, some poor fellow off at the edge of nowhere is going to be assigned an IPV6 address for his server farm. He then registers “mynewdomain.tld” and various subdomains.

OK … so now some client across the country on an older IPV4 network wants to browse to “www.mynewdomain.tld.” What happens then? DNS can’t return an IPV4 address equivalent for this new guy, because there are no more IPV4 addresses available!

That’s what’s puzzling me. I’m still fuzzy on how that’s going to work.

The ideal would obviously be to have everyone just switch over to IPV6, all at one time, one some date in the future, but I don’t see that happening.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

A bit more research may be in mind. The transition is made to be a
transition, not a cut-over. IPv6 is IPv4-backward-compatible. DNS can be
told to return A records (IPv4) or AAAA records (IPv6). You can see those
resolution requests happening today regularly. Whether or not something
is returned depends on if AAAA records are registered but when they are
not (as if often the case) applications fall back on IPv4 and all is well.
Get a LAN trace to see it in action even on your pure IPv4 networks.

Good luck.

On 11/05/2010 11:06 PM, smpoole7 wrote:
>
> cjcox;2248951 Wrote:
>>
>> I run our DNS… it speaks IPv6 fluently and handles IPv6 records just
>> fine.
>> So… not a problem. But as with ALL of IPv6, there’s a price for
>> handling
>> larger amounts of data. Most all routers/switches are optimized for
>> IPv4…
>> NOT for IPv6 (even the new ones with IPv6 support).
>>
>
> Well, heck, I’ll ask you, then. How IS the DNS transition going to be
> done? I’m thinking that eventually, when the available IPV4 addresses
> run out in the United States, some poor fellow off at the edge of
> nowhere is going to be assigned an IPV6 address for his server farm. He
> then registers “mynewdomain.tld” and various subdomains.
>
> OK … so now some client across the country on an older IPV4 network
> wants to browse to “www.mynewdomain.tld.” What happens then? DNS
> can’t return an IPV4 address equivalent for this new guy, because
> there are no more IPV4 addresses available!
>
> That’s what’s puzzling me. I’m still fuzzy on how that’s going to
> work.
>
> The ideal would obviously be to have everyone just switch over to IPV6,
> all at one time, one some date in the future, but I don’t see that
> happening.
>
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.15 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=6qb1
-----END PGP SIGNATURE-----

Ab, I understand the backwards compatibility, but that’s a one-way street. You can convert an IPV4 address to IPV6, but not vice-versa. How could you? IPV6 is 128 bits, while IPV4 is 32 bits. There’s no way to cram all of the info into 1/4 the space.

In my contrived scenario, the guy sitting at the edge of the Internet with a domain on an IPV6-only IP address is going to be out of reach for people who have IPV4-only ISPs.

Unless you’re assuming that Guy At The Edge has received both an IPV4 and IPV6 address? But in my scenario, I said that the ball had dropped, doomsday had arrived, and all available IPV4’s had been taken. That’s WHY he’s IPV6-only.

Ab,

To further explain (and this was my question to CJ … or to anyone else who has an idea of HOW the transition can be done smoothly):

  1. IPV4 doomsday arrives; the last address block has been eaten and there are no more IPV4 addresses available.

  2. Someone wants to create a new Web site – call it “doomsday.tld.” He is assigned an IPV6 static IP address (because that’s all that’s available now). He gets an AAAA record for that. But how are the IPV4 people going to find him with DNS?

If Doomsday Dood has IPV6 address F234:1234:A000:4567, that can’t be translated to an IPV4 address. So … when someone browsing on an IPV4 network does the DNS lookup, he’s going to get “domain not found” because there IS no “A” record for it, because there IS no IPV4 address. See what I’m saying?

For my money, if they’d asked me (and I’m HURT that they didn’t) (heh), I would have just added a 32- or 64-bit “country” or “region” code to IPV4 and been done with it. That would have expanded the available addresses by several orders of magnitude, with the only conflicts and questions arising when you try to move between regions.

But that’s just me. :slight_smile:

That was my question.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Two things. First… IPv6 is more than 4x bigger than IPv4. Yes, IPv4
has 1/4 the number of bits, but as a result it is significantly larger. I
imagine you understand this but just in case others are reading that is
worth pointing out. If my math is right then IPv6 is
79228162514264337593543950336 times bigger than IPv4 (that’s something
like 79 octillion times bigger, or really stinkin’ big).

In your contrived scenario somebody actually received an IPv6 address
without an IPv4 address for their website which is possible I suppose. My
first guess at how this will be handled is via tunneling. Getting an IPv6
address is simple enough on your private network and tunneling services
already exist so you could tunnel your IPv6 data to get to those places.

My personal opinion is that websites you want to go to will not be
IPv6-only for quite a while and by the time that bridge is crossed the
majority of us will have IPv6 addresses at least available to us. Maybe a
new market will become proxies that handle these types of situations,
though. When quests for A records are returned for a website that is
IPv6-only that site could have contracted with proxyA to handle that
request doing the tunneling transparently. Get in the market while it’s
hot; I haven’t patented the idea since it seems pretty obvious to me but
who knows if it will be necessary in the end.

Good luck.

On 11/06/2010 09:36 PM, smpoole7 wrote:
>
> ab@novell.com;2249015 Wrote:
>>
>> A bit more research may be in mind. The transition is made to be a
>> transition, not a cut-over. IPv6 is IPv4-backward-compatible. DNS can
>> be
>> told to return A records (IPv4) or AAAA records (IPv6). You can see
>> those
>> resolution requests happening today regularly. Whether or not
>> something
>> is returned depends on if AAAA records are registered but when they
>> are
>> not (as if often the case) applications fall back on IPv4 and all is
>> well.
>> Get a LAN trace to see it in action even on your pure IPv4 networks.
>>
>
> Ab, I understand the backwards compatibility, but that’s a one-way
> street. You can convert an IPV4 address to IPV6, but not vice-versa. How
> could you? IPV6 is 128 bits, while IPV4 is 32 bits. There’s no way to
> cram all of the info into 1/4 the space.
>
> In my contrived scenario, the guy sitting at the edge of the Internet
> with a domain on an IPV6-only IP address is going to be out of reach for
> people who have IPV4-only ISPs.
>
> Unless you’re assuming that Guy At The Edge has received both an IPV4
> and IPV6 address? But in my scenario, I said that the ball had dropped,
> doomsday had arrived, and all available IPV4’s had been taken. That’s
> WHY he’s IPV6-only.
>
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.15 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=T5So
-----END PGP SIGNATURE-----

Math error. It’s 2^96 times larger, not 4 times larger. Remember we are dealing with bits here.

In my contrived scenario, the guy sitting at the edge of the Internet with a domain on an IPV6-only IP address is going to be out of reach for people who have IPV4-only ISPs.

Unless you’re assuming that Guy At The Edge has received both an IPV4 and IPV6 address? But in my scenario, I said that the ball had dropped, doomsday had arrived, and all available IPV4’s had been taken. That’s WHY he’s IPV6-only.

Address allocations are not static, in fact addresses blocks get reclaimed. One way to allow servers to continue to have addresses is to migrate clients to IPv6. So far the number of servers is not comparable with the number of users.

You may find that your smartphone is already using an IPv6 address and working fine with IPv4 servers transparently.

OF course. I know that. I should have said, “IPV6 has 4 times as many BITS.” I knew what I was talking about, believe it or not. :slight_smile:

At the end of the day, there’s no way to cram a unique, huge 128-bit value into a slot that only has 32 bits. I guess I should have been clearer (though to be honest, I thought I was; obviously, I was wrong).

Address allocations are not static, in fact addresses blocks get reclaimed. One way to allow servers to continue to have addresses is to migrate clients to IPv6. So far the number of servers is not comparable with the number of users.

But my hypothetical was, what happens once this is no longer possible? In fact, going back to the math, once 2^32 IPV4 addresses have been assigned, allocated or otherwise used, then by definition, there are no more IVP4 addresses available.

I admit that it was a purely hypothetical question, and personally, I think the dire warnings that IPV4 address space will be exhausted next year are a bit premature, precisely because of what you just said. But asahypothetical, what if it did happen? Some guy calls his ISP, says, “I need a static IP address and I want to register a domain for a mail server” – and the only things available are an IPV6 address and an AAAA record in DNS.

The reason why IPV4 and IPV6 can live together at present (on my BlackBerry network, for example) is because there are IPV4 equivalents for all IPV6 addresses, or because the IPV6 subnet is being “translated” or “tunneled” into/out of IPV4 address space. If I have an isolated subnet, sure, I can make it IPV6 and then just carefully control access onto the IPV4 address space with a clever gateway.

It was strictly a hypothetical question. :slight_smile:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Just to be overly-specific, 2^32 addresses will never be assigned out.
Many of them (millions) are reserved for private networks. Many more are
reserved for multicast. More are reserved for weird test cases… etc.

:wink:

Good luck.

On 11/07/2010 09:06 AM, smpoole7 wrote:
>
> ken_yap;2249384 Wrote:
>> Math error. It’s 2^96 times larger, not 4 times larger. Remember we are
>> dealing with bits here.
>
> OF course. I know that. I should have said, “IPV6 has 4 times as many
> BITS.” I knew what I was talking about, believe it or not. :slight_smile:
>
> At the end of the day, there’s no way to cram a unique, 128-bit value
> into a slot that only has 32 bits. I guess I should have been clearer
> (though to be honest, I thought I was; obviously, I was wrong).
>
>
>
>>
>> Address allocations are not static, in fact addresses blocks get
>> reclaimed. One way to allow servers to continue to have addresses is to
>> migrate clients to IPv6. So far the number of servers is not comparable
>> with the number of users.
>
> But my hypothetical was, what happens once this is no longer possible?
> In fact, going back to the math, once 2^32 IPV4 addresses have been
> assigned, allocated or otherwise used, then by definition, there are no
> more IVP4 addresses available.
>
> I admit that it was a purely hypothetical question, and personally, I
> think the dire warnings that IPV4 address space will be exhausted next
> year are a bit premature, precisely because of what you just said. But
> asahypothetical, what if it did happen? Some guy calls his ISP,
> says, “I need a static IP address and I want to register a domain for a
> mail server” – and the only things available are an IPV6 address and an
> AAAA record in DNS.
>
> The reason why IPV4 and IPV6 can live together at present (on my
> BlackBerry network, for example) is because there are IPV4 equivalents
> for all IPV6 addresses, or because the IPV6 subnet is being
> “translated” or “tunneled” into/out of IPV4 address space. If I have an
> isolated subnet, sure, I can make it IPV6 and then just carefully
> control access onto the IPV4 address space with a clever gateway.
>
> It was strictly a hypothetical question. :slight_smile:
>
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.15 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=xfkF
-----END PGP SIGNATURE-----