OSDev.org

I was thinking of starting a small ISP in my area, in the UK, and im trying to find out who i need to contact in order to connect fibre optic cables from premises (FTTP) to my PoP. Google searches dont seem to show up anything. Do i need to get a permit from the council and hire a construction company to lay the cables? Or is there a company that can do connections between the office and customet premises for me? I dont know where else to ask, so if anyone at least knows a more appropriate forum that would be appreciated.

Although this is related to power cables, I would imagine the principles are the same: https://www.westernpower.co.uk/docs/con ... ts_v1.aspx

But surely you just lease a connection from BT, or similar, rather than providing your own cables.

how fast would such connections be? are they fibre optic?

You'd obviously need to talk to BT, or one of the other companies, but they would pull fibre to your premises if necessary. They already have the knowhow and resources to lay new cables. I suspect that trying to lay cables yourself would involve months, if not years, of negotiations and be extremely costly.

Another thing to consider is IP addressing. An ISP needs a large block of globally-routable IPv4 addresses, enough for one per customer, and if the ISP is planning to provide proper IPv6 support, a large enough IPv6 block for each customer to get a /64. You would also need a public BGP AS, which will cost you even more money...

Of course, there is a way to reduce the number of globally-routable IPv4 addresses you need: Carrier-Grade NAT. RFC 6598 specifies an address space, 100.64.0.0/10, for use in ISP deployments of Carrier-Grade NAT to effectively allow each piece of customer premises equipment to believe their public IP address is in the 100.64.0.0/10 network, while it actually gets overloaded to one of the handful of IPv4 addresses on the ISP's edge routers. Effectively, you are doing two network address translations -- one on the CPE and one on the ISP's CGN edge.

There are some pretty severe disadvantages to CGN though. You can't use static port address translation (aka port forwarding) on customer premises equipment, since the static translation will only apply to the 100.64.0.0/10 address, and not the ISP's actual globally-routable IPv4 addresses. Port Control Protocol may be able to be used to mitigate this, but if you only have, say, 14 globally-routable IPv4 addresses and a thousand customers, only 14 customers will be able to statically translate ports -- and you will likely need many of them for your own uses, eg. management.

You also have to be wary of malicious users causing your globally-routable IPv4 addresses to be banned. If you have 1000 customers, and 14 IPv4 addresses, and someone gets one of those globally-routable IPv4 addresses banned from Wikipedia, you will end up with a very large number of support tickets in your inbox.

Thankfully, IPv6 is a bit easier. RIPE requires prospective ISPs in their administrative region to be able to prove that they are at least multihomed before they are able to be assigned blocks of IPv6 addresses, but that requirement can get you a /48 (enough for about 2^16 customers). But then you have IPv6 connectivity for your customers, and probably no IPv4. At least they're not getting your entire provider banned from Wikipedia though!