OT: A broadband question.

It takes the same info in the router to rout a net whatever size it is.

Reply to
dennis
Loading thread data ...

Are you being deliberately obtuse, or what?

No, I think you really are obtuse.

Reply to
Bob Eager

Don't flatter dennis.

He's thicker than obtuse.

Reply to
The Natural Philosopher

Huh? how would that work. If ISP1 allocates fixed addresses to blogs engineering, then they can only keep them so long as they are with said ISP. Move to ISP2 and they won't be taking their fixed addresses with them.

Reply to
John Rumm

Just pointing out that your statement meant nothing. People that didn't understand how routing works might have thought you had a point.

And you statement wasn't obtuse?

Reply to
dennis

That is how it did work and people could take the addresses with them, and that is why we were in danger of running out of IPv4 addresses.

They had to try and reclaim addresses back from various companies and public bodies to put them back into bigger blocks. In the UK you can't get a "real static" IP address allocated to you any more (well not unless you are the government) all you get is an address allocated to a carrier and as you say they own them and keep them if you move.

If you can find the site that allows you to look at the route tables in the main USA network junctions (I can't recall what it is or find it on google ATM), you can look at the many tens of thousands of routes in there and see how many bits are routable, this tells you that there are still loads of small pools of addresses which can and do get moved around by people.

With IPv6 the intention is that only carriers will ever own addresses and route tables can be a few hundred lines rather than tens of thousands.

If you want to know more then look at BGP4, it is an attempt to reduce the number of route updates by confining the updates to smaller areas, even then it can take huge computers to handle all the updates in the core. At one time the core routers were using multiple Sun sparcs at each junction to handle the updates and were still being swamped if something broke. I expect they have even faster machines now.

Reply to
dennis

Wrong. Certainly not the ones I know. Security policy is often forced upon sysadmins from "above". These decision makers, it appears, are not technical.

Reply to
Mark

Mine goes down every night (and frequently more often) when I switch off the mains.

MBQ

Reply to
Man at B&Q

Hang on, you just said the administrators above forced the policy on the underlings, so you do know dumb administrators.

Reply to
dennis

There are very good reasons why a fixed IP address policy is a Good Idea.

If you are running a pro shop, anyone on a variable IP address is a noddy anyway.

You have to create a whole new level of security to deal with it, at huge expense and loss of simplicity and increase of vulnerability.

Why failure to do so would be considered 'Dumb' is beyond me.

Reply to
The Natural Philosopher

Actually two or three hundred thousand by now, I should think. Here's another useful link you may wish to peruse:

formatting link
With IPv6 the intention is that only carriers will ever own addresses and

Routing updates are handled by the routers themselves. There's also something called "route flap damping" which reduces the propagation of updates so that if Joe Soap's link is going up and down ("flapping") because of some intermittent fault, he's not causing a continuous stream of routing updates all over the planet.

Reply to
Tim Streater

Indeed. Last time I set up a core router, a big cisco was well capable of handling the full Internet routing table.

Updates are in fact quite slow.

It takes the best part of several hours to acquire the tables though.

Reply to
The Natural Philosopher

Oh dear someone that thinks a static IP address actually provides security.

I prefer to use something that does provide security like hardware tokens and one time use passwords to enable a tunnel. then it doesn't matter what IP address the trusted user is on. Wow! the user could even log in from a public Wi-Fi access point or a mobile phone.

Come to think of it many people do, so they can't be using static addresses.

I expect TNP falls into the dumb administrator category.

Reply to
dennis

Not all routers do their own route calculations. It depends on where they are and how many routes they are handling. I suppose that an attached multiprocessor machine can be a part of the router but that depends on how you choose to view a router, e.g. as a route engine and an administration engine, the two being effectively independent. The two items do not have to come from the same manufacturer either.

BTW I have actually used some biggish routers and ATM switches (multiple

80Gb per second switch matrices) and I know what is inside them.
Reply to
dennis

For anyone actually interested in IP routing..

telnet route-views.oregon-ix.net

This is a cisco router where you can actually enter commands like "sh ip bgp" and examine some real internet routing tables.

Reply to
dennis

The routers I have dealt with (large Junipers at a million quid or so loaded up) have primary and redundant routing engines built in. I don't think they'd want or expect their customers to use third party devices to handle the routing updates & I never heard of Cisco doing it either.

The actual packet-forwarding and route lookup in the Junipers was done effectively in hardware, along with packet filtering. They could handle any amount of complex filtering of packets at wire speed (wire speed being 10Gbps and possibly faster these days). Impressive devices.

Reply to
Tim Streater

Even if you don't turn it off. its not unknown for BT to lob you off from time to time in the dead of night as they fiddle with their kit.

Reply to
John Rumm

Yes I have used them before. The ones I used were near a 1/2 million pound a card, but they weren't Juniper or Cisco.

They can't really cope well with suddenly having all the routes change at once as can happen when a major node goes down. They can take hours to rebuild the tables and that isn't limited by the speed the info arrives, route exchange isn't done in periods of hours.

IIRC not all routing is done in hardware, the hardware CAM is a bit small for a big router.

Have you telneted into route-views.oregon-ix.net yet? You can see just how many /24 addresses have to be routed and its not a particularly big table.

Reply to
dennis

:-)

No I used to do it on our routers occasionally but gave all that up when I retired two years ago. Certainly there'll be a lot of /24 left over from the old days. We only had a small numbers of large customers (each was a national research network such as UKERNA/JANET or DFN) and so since they tended to keep their RIPE entries up to date our part of the routing table could be precisely derived from the RIPE database.

The DANTE networks are also entirely IPv6 ready, but it was interesting (at least at the time 3 - 4 years ago) that the commercial upstream providers that DANTE used had difficulty providing a proper and general IPv6 service. They kept saying "Not much call for that round here, Squire".

Reply to
Tim Streater

HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.