What is wrong with IPv4?

[#1]

In this sub forum, there are many questions as to when IPv6 will be coming to plus.net - so I decided to ask the question from a different perspective, what are users problems with IPv4?

Now want the question to be used focused, e.g. the common reason given for IPv6 is that IPv4 addresses are running out. However I haven't heard of any plus.net customers not able to get an IPv4 number, or being on CGNAT. So IPv4 addresses running out isn't a problem for plus.net as far as we can tell, so not a problem for users.

So instead, what are your problems with IPv4?

For me the major problem is being behind a NAT on my home router. Firstly setting up a NAT is a hassle; but more importantly its difficult from the WAN to connect to anything at home. E.g. if I wanted a simple web server on a home computer, I need to set up a specific DNAT on my home router - and this is a hassle. This has to be done for each connection.

I don't have problems connecting to IPv6 web sites - as I run 6in4 so can connect. However all 6in4 packets are routed via a specific address, and IIRC this means that plus.net makes all 6in4 packets go via he.net. This hasn't caused a problem but could it cause a problem?

So what are other users experiences? How has being stuck on IPv4  affected your internet connection? What impact has it had?

[#2]

You are right that IPv4 isn't a problem for most Plusnet users - most probably don't know what IPv4 or IPv6 is, they just want the internet (or rather specific applications) to work for them.

IPv4 addresses are limited, but in fact there are more IPv4 addresses available now than there were a few years ago - because a lot of large companies have moved to IPv6 addresses on internal networks and have sold their IPv4 ranges.

However IPv6 is desirable for several reasons, including:

- future proof (for the foreseeable future)

- faster routing (not true with IPv4 to IPv6 bodges)

- lots of addresses so IoT devices can have unique public addresses without ugly port-forwarding

I have IPv6 on my Plusnet account by using a L2TP tunnel to AAISP - I get AAISP's fixed IP addresses (IPv4 and IPV6) etc, but pay Plusnet prices.

[#3]

@corringham yes I think both our thoughts are in agreement.

1. future proofing to me is giving enough IP addresses for the world, this though is a problem for plusnet, and not us users - so this is pressure on them not us.
2. faster routing, whilst in principle I agree, as 6in4 and 6to4 for plusnet go via he.net. I'm not sure what real difference it gives. e.g. a tracepath gives: vl241.ptw-ag02.plus.net: 49.815ms, 192.88.99.1 58.667ms; and a ping to 192.88.99.1 as 36.3ms. So most of my delay is going across an interlaced ppp connection to my plusnet gateway.
3. Yes, same as my point about running web servers, or external ssh connections, etc

So what I see here, is that most of the pressure to move to IPv6 is upsteam in plusnet. And as far as us users can see, this isn't something that plusnet worry about at the moment. So from a user position all we can do is wait and see what happens. Maybe just confirm our adsl routers are IPv6 compatible.

The above though does make me wonder what plus,net relationship with he.net is? E.g. at the moment they route 192.88.99.1 to he.net. Is this a formal agreement? If it isn't how long will he.net tolerate this?

[#4]

Aside from nothing being wrong with IPv4, there are some reasons I would definitely want IPv6 and hopefully say goodbye to dual-stack with IPv4 as well. Generally I don't want to have NAT because it's not without it's costs, and I'm looking forward to hopefully lower latency and better throughput if IPv6 ever goes fully native on service providers. Of course, IPv6 packets are more effecient as well so there's that. I'm generally just looking forward to better effeciency across the board. On the other hand, IPv6 is not as easy to understand for people because of the hex addresses and personally I struggle with them and their subnets, aside from knowing what the defaults usually are and how to shorten them. My personal opinion is if IPv6 had been in dotted decimal a lot more people would be asking for their providers to get with it and just do it already.

[#5]

I would have thought that the increased size of the IPv6 headers, with the maximum packet size remaining the same (limited to the 1500 bytes of Ethernet), would make IPv6 less efficient in terms of throughput.

[#6]

The packet header is simpler (unless extension headers are used I assume) and there is no need to checksum the packets at each router. But my post might be slightly misleading, I'm mostly thinking of NAT and overhead for processing packets. The good thing about NAT provides easy security even if it's not designed for it.

But still I can't get my head round why 640 undecillion IP addresses are needed or even comprehend that number, I would have thought a few trillion would do it. Human factor was not considered maybe. I suspect with native IPv6 on my home broadband I would 'tolerate' some of the features rather than liking them.

I think most ISPs are going to hang in there with tunneling, and then go dual stack only when forced to.

[#7]

IIRC one advantage in IPv6 is that the hop count/TTL field is outside of the part of the packet that is covered by the checksum - while in IPv4 the checksum needs re-calculating after each hop. I could be wrong on that though.

But I disagree about ISPs hanging on with tunneling. Some of them are, but many have already gone dual-stack or even IPv6 only with IPv4 at the client handled with (eg) NAT464 or 4 in 6 tunnelling. But how it's handled in the ISPs network is generally of no concern to the user (it is if you use your own router) - all most users are bothered about is what they see on the LAN side of the router. Many of the smaller ISPs (eg Zen, A&A) embraced IPv6 a long time ago; and both BT and Sky (2 of the big 3, don't know about StalkStalk) enabled IPv6 some time ago.

[#8]

I sometimes have mixed feeling about NAT, on the one hand some things down't work well behind a NAT; and routing as a whole gets easier with IPv6.

However that said if we convert to IPv6 and no NAT, then plusnet through recording packet orgins could trace how many machines I have. Not this shouldn't change my bandwidth; but when they know I have 8 or so devices that connect to the internet - would they question if that should be the same price as someone with only two? As more IOT come along, this number grows (e.g. am looking at getting central heating installed - and would like to get that on my local network, so I can both monitor and switch on and off remotely).

Trying to remember what I read about IPv6 headers, yes as I recall it structure was very different from IPv4, and yes IIRC there were far less options with IPv6 - mainly getting ride of the stuff not used on IPv4. Can't recall though how size compared.

[#9]

These days the use of fixed IPv6 addresses (using ethernet MAC address derived addresses) are long since deprecated. The expectation now is for nodes to use multiple addresses and change addresses over time. So Plusnet cannot know whether the (say) 8 addresses it sees are 8 addresses on one device, or single addresses on 8 different devices.

[#10]

Simon,

Good point - and actually had I reflected I'd have known that on my 6in4 I can see that IPv6 address regularly change, as (under linux) it keeps the old adress in case more packets sent - so each IPv6 node ends up with 10+ addresses. Looks like the old ones are marked "scope global temporary dynamic" whilst the new ones are "scope global mngtmpaddr noprefixroute"

[#11]

Well, just a case of seeing if plusnet will sort this out. I wouldn't move provider over it exactly, but I'm not always sure if I'll continue a contract with an ISP until I've seen them in action for a while and I'm relatively new with PN. So wether I'm on IPv6 or not comes down to what I feel like at the time I suppose.

No announcements on progress yet, but in the end I think dual-stack will be needed for some time for programs that just won't get updated wether ISPs offer native or not.

[#12]

@SimonHobson 

I knew sky had done that, but had no idea BT had completed that work yet. So the situation has changed a little since I last checked it.

VM have been dragging their heels for a long time and frustrating users by refusing to give them any information about it, but a quick check just now shows something may be different and it looks like they have been trialling it since last year.

I know not so long ago IPv4 just run out completely, perhaps that's why they're getting it together. There might be a few more addresses in Africa but that's not helpful to us, so maybe they are finally moving things at the ISP end.

I'd say you have the right of that then, maybe some ISPs will drag their heels a little like they have been for a while but it looks like this will happen soon if major ISPs are doing it.