[good8675309]

Ethernet at 50 and still not slowing down - a testament to the saying 'if it ain't broke, don't fix it.' It's remarkable how this old-school tech has remained relevant in an age of wireless everything. Makes me think of my tech stack, Java, monolith, sql db, etc. and it keeps serving all of my paying customers reliably, even though I'm not using the latest everything.

[fanf2]

Ethernet has been redone from scratch several times. What has remained fairly constant is its frame and address layouts - an example of Kleinrock’s “narrow waist” pattern in protocol design.

[kibwen]

It's honestly impressive how few qualms people have with the link layer, considering all the kerfuffle that gets raised for the address layer (IPv6) and the transport layer (QUIC) and the application layer (HTTP2/3). It's nice to have at least one protocol that feels "finished" (in a good sense).

[projektfu]

I think the only part that hasn't been "fixed" is that it sends frames. 10Base-T is a new topology and cable with the bus virtualized in a hub. Switched Ethernet is like a different, intelligent network larping as Ethernet to naive stations. Signaling changes as you increase data rates and media now stretches from copper to fiber and the standard has retired baseband coax above 10Mbps. It's cool that they've made it work and kept it under the same "roof" of standards, so that we don't have to make so many vendor decisions or have regret building out one vendor's network and watching another one with different cabling or signaling take off with higher speeds and lower prices.

[msla]

> I think the only part that hasn't been "fixed" is that it sends frames.

And this is just terminology: An Ethernet protocol data unit is a frame because it's Ethernet and an IP protocol data unit is a packet because it's IP. You could switch the terms around and nothing would actually change.

[floxy]

"I don't know what language will be used to program high performance computers in 50 years, but we know it will be called Fortran."