Re: [apps-discuss] IRC URIs to denote users

[Sean Leonard <[e-mail]>]

On 9/22/2014 1:50 PM, Melvin Carvalho wrote:
> I was wondering if IRC URIs are standardized at all?
>
> The latest I found was :
>
> http://tools.ietf.org/html/draft-butcher-irc-url-04
>
> I have a use case of marking reputation from one system (web chat 
> room) to an IRC chat room, but I require an identifier for the user.
>
> The suggestions so far have been:
>
> irc://user@host
> irc://user@host/ -- trailing slash
> irc:user@host -- similar to xmpp
> irc://host/#user -- fragment could be problematic as per RFC 9386
>
> I've gone with the second option for the moment
>
> Any pointers would be most welcome.

Based on the options presented, I think the third one, irc:user@host, 
best captures the spirit of the URI syntax.

RFC 3986 Section 3 and 4.3 define absolute URIs as scheme : hier-part; 
hier-part can be one of:
"//" authority path-abempty
path-absolute
path-rootless
path-empty

The first one is usually used for identifying resources accessible via 
some Internet-related protocol where the authority is a host:port (plus 
optional userinfo to log in to the host:port), followed by some 
server-specific path to the resource (usually hierarchical). The second 
one is for some server-specific path (usually hierarchical), where the 
server's host:port are "obvious". For example, 
<ldap:///o=University%20of%20Michigan,c=US> means "an LDAP URL referring 
to the University of Michigan entry, available from an LDAP server of 
the client's choosing" (RFC 4516 Section 4). <file:///> URIs (URLs) 
refer to the local file system.

The fourth one, path-empty, isn't used much...but the gist (I suppose) 
is if you want to skip right to the ? query or # fragment parts. magnet: 
(provisional) URIs use this.

Thus, we arrive at path-rootless, which is basically "Everything Else". 
In this case, you are trying basically to identify an IRC object (a 
user), and tag it in such a way that it's not an e-mail address. 
<[e-mail]> looks like an e-mail address. <irc:[e-mail]> 
looks the best to me; it's obviously not an e-mail address. Similarly, 
it's obviously not an address to an IRC channel. If you say 
irc://host/user, you're sharing the channel namespace with the user 
namespace...not a great idea. If you say irc://host/#user, you are 
saying that the path is "/", which (in the context of users/nicknames) 
doesn't make a whole lot of sense since users/nicknames are specific to 
servers/hosts, not to sub-parts of servers/hosts.

The telnet: URL/URI is the most analogous to irc URIs--and telnet URLs 
go way back to Tim Berners-Lee's [original paper]. In the original, it's 
telnet://userinfo@host:port -- with no trailing slash (i.e., authority 
path-abempty). In RFC 1738 sec. 3.8, it's telnet://userinfo@host:port/ 
-- with trailing slash (i.e., authority path-abempty).

The ssh URI proposal (draft-ietf-secsh-scp-sftp-ssh-uri-04) has no 
trailing slash...but it also says that the path part is irrelevant. In 
any event, the overall gist is that "//" at the beginning is sufficient 
to distinguish "network paths" from "other things".

Sean

[original paper]: 
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.45.1836 
"Universal Document Identifiers on the Network"