SETNAME Spec

[justjanne]

In Short

• Introduces a new setname capability
• Introduces a new bidirectional SETNAME command

Rationale

Realnames have historically been immutable for a connection, which has
historically been accepted, as they were rarely used and usually not as present
in client UIs. Still, multiple IRCds have provided independent, non-standard
commands to update realnames.

Nowadays, multiple clients show realnames directly in chat, next to the nick
name, or use information from the realname to provide enhancements, such as
using emails in the realname to provide gravatar based avatars, leading to
higher demand for this functionality than ever.

[DanielOaks]

Nice, I like it.

The draft/setname capability. I presume that the s2c SETNAME message will only be sent to clients that have negotiated that capability? (and clients that have not negotiated it will not receive the SETNAME message).

With servers validating the new realname, if a given realname is too long it feels like the server should probably just silently truncate it, matching what servers already do when getting it initially via USER.

If the server would otherwise reject the change (maybe because the client's been changing it too fast, it's being used to bypass a channel ban like how nick changes are restricted, or the value's just bad), it might make sense to explicitly send a response. Maybe a good chance to use the proposed FAIL message, something like this:

FAIL SETNAME CANNOT_CHANGE_REALNAME :Slow down your realname changes
FAIL SETNAME CANNOT_CHANGE_REALNAME :Cannot change realname while banned from a channel
FAIL SETNAME INVALID_REALNAME :Realname is not valid

[jesopo]

Very simple implementation in BitBot as of bitbot-irc/bitbot@acbe02ba - let me know when there's a test server to try it on

[DanielOaks]

PR written up for Oragono here, pending the review q above right now it fails if the sending user hasn't negotiated the cap Updated the implementation to match intended behaviour, the spec language should explicitly state the intended behaviour.

[jwheare]

Should there be a NAMELEN isupport key?

[DanielOaks]

On how existing impls handle the realname being too long in their current SETNAME commands:

Insp:

user->WriteNotice("*** SETNAME: Real name is too long");
maxreal is 128 by default

Unreal:

sendnotice(sptr, "*** /SetName Error: \"Real names\" may maximum be %i characters of length", REALLEN);
REALLEN is 50

So they both explicitly send a notice to the user and fail out, not changing the realname. Cool.

[Zarthus]

Does it make sense to call this SETNAME over SETREALNAME or SETGECKOS?

When I think of setname I think of /nick, personally. Alternatively to be consistent with /nick you could just call it /realname too

To add on Dan's point: charybdis silently truncates realnames at 32/64 characters I think.

[DanielOaks]

/SETNAME is already in use by Insp and Unreal, if this could be compatible and use the same command name that'd be cool. If we would invent our own from scratch though I'd def go for /REALNAME to be consistent with /NICK ;)

[jesopo]

+1 for the command to be REALNAME not SETNAME

[jwheare]

I'd favour not changing a command that users might be accustomed to just for the sake of it.

[jwheare]

This is now deployed on the IRCCloud client, and our teams/slack/bnc servers. @jesopo you should already have access to the testing team server.

[DanielOaks]

Looks good to me, nice work.

[jwheare]

Some change suggestions.

[jwheare]

Thanks, this is mergeable as draft now imo.