Distinguishing servernames from bare nicknames in source prefixes

[jwodder]

If a client receives a message whose source does not contain either ! or @, are there any rules that could be included in this spec for determining whether the source is a servername versus just a "bare" nickname?

• I assume that servernames are supposed to be internet hosts (i.e., a domain name or IP address), but, as far as I'm aware, there's no rule against a nickname also being a valid host (though https://modern.ircdocs.horse/#clients states that nicknames "SHOULD NOT contain any dot character").

• At first, I thought that servername prefixes are supposed to always equal the name of the server that the client is connected to, and that they remain the same throughout a connection, but this would make servername prefixes pretty much superfluous, and the statement that "The source indicates the true origin of a message" somewhat suggests that a source could be the name of another server in the network.

[emersion]

There is no reliable way to know. The convention is to use names with a dot for server names, and names without dots for nicknames. But it's just a convention.

[SadieCat]

I've considered adding a vendor tag for differentiating messages from servers but without universal support it doesn't really make sense.

[emersion]

I believe it would still help if there is a cap guaranteeing that all server messages will carry the tag. Clients could disable their heuristic in that case.

[tommyrot]

The wording in this repo regarding nick and servernames is weak unfortunately. In reality servernames must contain a dot and nicknames must not contain a dot.

See also related discussion in #148

As for your second point: some commands allow a user to request information from a different server on the network as one is connected to, thus the reply will indeed be of that other server.