I've been doing some reading about MTU and I just can't get my head around something. When we talk about MTU for ethernet, it's documentated that the MTU is 1500, however the frame size is larger after the Ethernet headers are added. So effectively, when people talk about the MTU for ethernet being 1500, they are referring to the payload.
Now then, I could accept the above statement if this was a standard case for every other protocol. But with TCP for example, the MTU is considered the payload + the IP header + the TCP header. If you compare this to Ethernet, the MTU was just the payload, but with TCP it's the payload + the headers.
Am I understanding this right, because this is really confusing to work out which protocols include the headers with the MTU, and which don't? I have a sneaking suspicion that it's just Ethernet that does not include the headers when talking about MTU, since it would be setting the standard for the maximum packet size for upper layer protocols (i.e. you could only ever have an IP packet with a total packet size that uses a 1500MTU including all the headers and everything). But I want to get some clarity on this topic.
I do realise that fragmentation can be used, but I want to know how to work this out without talking about fragmentation.