Hacker News new | past | comments | ask | show | jobs | submit login

BGP is the routing protocol of the Internet. There effectively is no other choice of routing protocol between autonomous systems. A reasonable synonym for "Internet" is "the global BGP routing table".

BGP also doesn't use multicast, you may be thinking of OSPF on multiaccess networks. BGP uses tcp/179 unicast to the IP addresses of its configured peers.

That said, multicast works just fine over the Internet. It's not commonly used, certainly not by home users and not very often by enterprise users, and was phased out on Internet2 by 2021 (I think?), but there's absolutely nothing in principle that would make it not work.




> That said, multicast works just fine over the Internet. It's not commonly used, certainly not by home users and not very often by enterprise users, and was phased out on Internet2 by 2021 (I think?), but there's absolutely nothing in principle that would make it not work.

In principle, no. In practice, I don't think many ISPs have equipment configured to forward multicast, except for those using multicast for TV and those probably don't interconnect with others.


Many years ago UK ISPs particpated in an MBone with BBC, ITV etc providing live broadcast

https://www.bbc.co.uk/multicast/tv/channels.shtml

Brandon Butterworths note about "why"

https://support.bbc.co.uk/multicast/why.html

Shows the growth of the backbone and CDNs:

> The Olympic audience is expected to be around 50K streams, delivering 10Gbit+ is on the limit of sensible unicast delivery.

In 2020 the BBC's internal CDN was delivering 100 times that [0] for 250k users, and 5 years later I suspect it's another order of magnitude given that iplayer does 5 million concurrent live views quite frequently [1,2]

[0] https://medium.com/bbc-product-technology/bbc-online-2020-in...

[1] https://www.bbc.co.uk/mediacentre/2024/audiences-flock-to-bb...

[2] https://www.bbc.co.uk/mediacentre/2022/england-v-iran-bbc-li...

By 2035 and TV turnoff there's no reason to believe that the infrastructure won't have been able to scale another 100 fold and handle 500 million concurrent live streams. Makes no sense to multicast out 30 different formats, to people on phones and tablets and TVs hanging off wifi. It's a very different consumer experience than a PC wired into an ISP like it was in 2007.


You can definitely route multicast over the Internet via some kind of tunnel, be it stunnel, VPN, GRE. The ICE exchange uses stunnel for dev/certification multicast.


FWIW, 100% that BGP itself doesn't *use* multicast, but it can *propagate* multicast routing information. It's certainly technically possible to support multicast on the Internet (..thus the invention of MBGP) but in practice has been a non-starter for a whole bunch of reasons.


It's sort of correct and sort of incorrect to say it's the only possible routing protocol between autonomous systems. Since many inter-AS connections are effectively made-to-order, they could use a different protocol, but in order to participate in the global BGP mesh, it would have to have semantics similar enough to BGP. Most notably, it would have to support the concept of AS-path. However I don't think there are any universal requirements beyond that. BGP isn't a single distributed algorithm like say OSPF - a completely separate instance of BGP runs between every pair of connected ASes and they share some data, indirectly forming a global system.


Anycast is pretty useful on the Internet :)


Anycast is a very different beast, though. Anycast is just unicast but you announce the same IP space from multiple destinations, and the network figures out how to get a packet to the closest one. If one of those destinations fails, it just goes to the next closest one.

Unicasts, multicasts, and broadcasts all actually work differently underneath and require specific handling by network equipment. Anycast is just a special case of unicast and generally speaking network equipment is completely unaware of it.


In fact, IPv4 and IPv6 both have a reserved range of multicast addresses (formerly called "Class D"):

https://en.wikipedia.org/wiki/Multicast_address

There were more than a few people who spotted how disused this range had become after mbone experiments, and sometimes suggested reclaiming the range as IPv4 address space was being exhausted.

Interestingly, there are reserved multicast addresses (yes, addresses, not ports) not only for OSPF, but for many other interior routing protocols, as well as mDNS, LLMNR, and NTP. Conspicuously absent is any reservation for BGP.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: