RTFM problem.

[kroberts]

Hi,

I have read good-sized chunks of IPV6 Essentials, seen here:  http://search.barnesandnoble.com/IPv6-Essentials/Silvia-Hagen/e/9780596100582/?itm=1  This book describes the technicalities of packets, but not so much how to set up your network in a small office.

I've also browsed around IPV6.org and a few other sites, including a bunch of stuff on http://he.net.  I've worked with Cisco firewalls, routers and managed switches, all of which in its current version of our software support IPV6.  I've compiled, configured and used Linux firewalls in a production environment using IPV4, although that memory is very rusty and outdated.  I configured our Cisco gear with VLANs, DHCP forwarding, all that.  I'm by no means a Cisco certified anything at all, but I'm not a beginner either.

All I can say is, I'm still very very lost.

I'm an application developer and have been since the 80's.  I have not yet tried to mess with IPV6 because it's a non-billable item for me, and frankly I don't feel that I have a good enough handle on it to start playing.

I would love to RTFM, but I can't find the FM to R.  Everything seems outdated or seems to be written either for a more beginning audience than I am (e.g. the intro on this site), or a more advanced one (e.g. the book above).

My confusion is in part because of language.  The book mentioned above seems to be mostly written for an ISP audience.  I'm a software developer, who happens to run the network in a small company.  I want our apps to work with IPV6 so I feel I need to set something up.  We generally don't code to TCP directly, but I'm in the financial sector so we need to have tested it before we can say it works.

My hope is that when I ask a few questions somebody will point me to the book or web site that can put me over the edge into comprehension.

First, a technical question.  What exactly is "link layer?"  I would at first call it a subnet, but that doesn't seem to cover it.   I would call it on par with ethernet, but that also doesn't seem to be it.  All the other levels of scope seem straightforward, just not that one.

Second, allocations:
According to the above book, a single organization gets a /48 address by default.  This seems totally bizarre.  HE is offering /64 addresses through the tunnel broker, which in my understanding maps to a single non-divisible subnet.  Or a /48 address for pay customers, which is so immense an address space that I can only think they mean to an ISP for further division.

I do understand the theory about using the MAC address of the card as an identifier and placing it directly into the address.  I understand the sparse allocation scheme.

What I want to find out is if there is a generic, workable plan for a small business to get on IPV6 that covers all the bases of experimentation, conversion and implementation?

The good part is that most of our server-side software seems to be IPV6 enabled, and most of our workstations seem to have an IPV6 address even though the switches are ignoring them.

In case it becomes interesting later, the Cisco gear is 2600/515e/3550.

If someone could recommend some more reading that seems suited to my needs, I would greatly appreciate it.  It can be a book that I buy, or a web site.

Thanks.

[ericj]

I don't really have any books to suggest, but I'll try to answer you questions.

First, a technical question.  What exactly is "link layer?"  I would at first call it a subnet, but that doesn't seem to cover it.   I would call it on par with ethernet, but that also doesn't seem to be it.  All the other levels of scope seem straightforward, just not that one.

The link layer is what carries network layer (such as IP) packets. Ethernet is one example. Tunnels can also be seen a link layer, since the IP packets are carried over the tunnel.

Second, allocations:
According to the above book, a single organization gets a /48 address by default.  This seems totally bizarre.  HE is offering /64 addresses through the tunnel broker, which in my understanding maps to a single non-divisible subnet.  Or a /48 address for pay customers, which is so immense an address space that I can only think they mean to an ISP for further division.

First of all, you don't need to pay to get a /48 tunnel from HE. You can allocate a /48 for you tunnel if you need to.

I agree that /48 is an enormous amount of address space. However, their are a very large amount of /48 blocks available. It would be possible to allocate one IPv6 /48 to every single possible IPv4 address, and it would only use up a single IPv6 /32. (In fact, 6to4 does exactly that, using 2002::/32.)

Giving businesses /48s or even just /56s instead of /64s allows them more freedom to divide their network as they feel fit.

[kroberts]

Eric,

So you're saying that "link layer" is the same old just-above-the-hardware layer that it was in the old model, only with a few new constructs?

That would make things easier to understand, only those changes are going to be tough to swallow.


It's funny, I did the math on the subnet size but didn't do it on the number of /48's available.

I guess that more or less answers those questions, but I would still like to see some sort of plan for upgrades, or a real case study for an average small business before diving in.

Call me a coward.

Thanks though.   :)