OT: A broadband question.

[John Rumm]

There are a number of issues here. Firstly one thing that many ISPs need to deal with is growth. Hence they will often need to add backhaul capacity as their user base grows. Being able to shift customers around between pipes, and connection points in their networks is handy. Static IPs can make this harder.

There are also a number of customers who in effect "dial up" to broadband. Typically anyone using a USB modem. Hence their connection goes down every time the computer does.

Yup, also if you need to maintain a secure server remotely, then being able to lock down the IP addresses that it will accept incoming connections from can greatly aid security.

[John Rumm]

The fears that we would run out of addresses "soon" have largely been displaced by the widespread adoption of NAT these days.

To describe the most successful and widely used addressing scheme as a c*ck up seems a little less than charitable. Especially when you consider how resilient it has proved in the face of growth to many magnitude larger than even the wildest predictions and dreams of its creators.

Whilst one can be wise in hindsight, its interesting to note that many of the naysayers were on the planet at the time, and could have had input to the design, but didn't.

I can see IPV4 being with us for a substantial amount of time yet.

or good ones - depends on the circumstances.

[Bob Eager]

That would be why my ISP has alllocated me a static /48 block, then.

[dennis]

If being a very big if.

[dennis]

There are and always have been loads of IPv4 addresses available, AFAIK there isn't a shortage now. What there is is a limit to the size of router tables. If you are stupid enough to allocate 16 IPv4 addresses to say bloggs engineering and they move providers to BT then that block of addresses needs a router entry in all the global routers that want to talk to it. Do that for a few million customers and the result is what we have now, huge router tables that take ages to propagate and are error prone. Its also totally unnecessary, DNS is designed to avoid the need and nobody needs to address using IP addresses rather than DNS not even firewall administrators. When nobody uses IP addresses then being static is irrelevant and all addresses can be hierarchical. The router tables drop to a few hundred entries, everything works faster, the network reroutes faster, everyone wins except Cisco who can't sell their big router engines anymore.

Dumb administrators, period.

[dennis]

If you change ISP can you take those addresses with you?

[Tim Streater]

That's rubbish. No ISP does that and no ISP will route 16 addresses belonging to another ISP. In fact, that is not technically possible. The original 16 addresses will have been allocated by RIPE (if we're talking about Europe) to the first ISP.

The 16 will be part of a larger block, and that larger block will be a single router table entry that stays with the first ISP. ISPs build their router tables from the RIPE (and other) databases; the second ISP would be unable to route the 16 addresses, even if they wanted to.

The exception to this is if Bloggs engineering was an ISP themselves, but then they would be their own ISP with their own block of addresses allocated by RIPE.

I know whereof I speak in this matter so lets not have any shit.

[Bob Eager]

But they don't. You move ISP, you lose the addresses. That's always the case.

[Bob Eager]

No, of course not. For exactly the reasons you mentioned. 'Static' doesn't mean 'forever' - or are you wriggling once again?

[Andy Burns]

Early IP alloctions are what is now known as "PI" provider independant, e.g. a class C that was provided by the JIPS NOSC nearly 20 years ago

It's not been routed for over 10 years, I don't know if ISPs will tag other people's subnets to their own AS nowadays, but that was the arrangement we had at the time.

[Tim Streater]

Well, PI space is rather harder to come by now anyway, for just this sort of reason. They do try and get everyone to aggregate, that's the way to reduce the table size. Also, if the space hasn't been routed for

10 years RIPE may start asking for it back eventually. We tried to be good boys at DANTE and returned space and even AS numbers.

[The Natural Philosopher]

So do I and da man spik da troof.

ISPS have large aggregated blocks: These are waht are held in the core routers of the (BGP) world. Once inside the ISP they will then use finer graduations to route to the actual destinations.

Only if you connect to two ISP's will you need to advertise BGP to your upstream providers and have your actual range of addresses advertised outside your upstream provider(s).

[The Natural Philosopher]

Not always

If you have your own block.

And its big enough.

Its not applicable to domestic broadband, but it is to large corporates.

[Andy Burns]

Oh I'm sure it is, but back then, Class C subnets were handed out for the price of a faxed document.

[dennis]

There are millions of addresses where that is not the case. Addresses were handed out to companies and individuals. There has been a lot of effort to recover some of those addresses but not everyone cooperates.

Look at

and see which ISP has assigned the addresses and which one can get them back.

Hint "direct assignment" is the clue here. You can easily find many allocations to none ISPs

[The Natural Philosopher]

The theory being that if you knew what one was, you probably needed it.

[dennis]

Oh look its the same Tim that said I was talking shit now talking about exactly what I said.

[Tim Streater]

I tole you, bozo, that anyone with 16 addresses won't be able to move and take them with them; they'll be told to renumber.

And global ISPs, these days, the likes of Global Crossing, COLT, Telia, and so on, won't route anything smaller than a /24 anyway.

[Tim Streater]

For some info about the fact that IPv4 space is running out see:

[Bob Eager]

That was what I meant. If it's a big block there is no routing issue.

Mind, I'm domestic and I have a /26..