[Uri-review] Re: [art] Re: Alternative representation of URIs in YANG

[Kent Watsen <[e-mail]>]

> On Dec 8, 2025, at 8:54 PM, Dale R. Worley <[e-mail]> wrote:
> 
> But referring to my earlier message, I don't think the proposed module
> has that problem.  

I hope this is true.  Though folks are stating URIs are nearly impossible to structure, RFC 3986 (which is not obsolete) does not indicate this, and the only "update" to it (via RFC 8820) doesn't seem to indicate otherwise either. 


> Maybe I've overlooked something in the discussion.
> Or maybe I've overlooked that the module has "host" as being mandatory,
> which for URIs it certainly isn't.
> 
> The overall syntax of (absolute) URIs is
> 
>      URI         = scheme ":" hier-part [ "?" query ] [ "#" fragment ]
> 
>      hier-part   = "//" authority path-abempty
>                  / path-absolute
>                  / path-rootless
>                  / path-empty
> 
> so if you know the scheme, the authority (which may be empty), the path
> (which may be path-abempty, path-absolute, path-rootless, or
> path-empty), the query (which may be empty or absent), and the fragment
> (which may be empty or missing), you know the URI.

I think folks are misunderstanding the "presence" statement.   Also, please note that, in YANG, the default "mandatory" value is "false".  Lastly, the default "string" type in YANG may be empty.  Together, the following abridged YANG:

    container authority {
       presence
         "Indicates that 'authority' has been configured.";
       leaf userinfo {
         type string;
       }
       leaf host {
         type inet:host;
         mandatory true;
       }
       leaf port {
         type inet:port-number;
       }
     }

Should be read as:

The "authority" node MAY be configured.
If the "authority" node is configured:
the "userinfo" node MAY be configured and, if configured, MAY be empty.
the "host" node MUST be configured.
the "port" node MAY be configured.



> Looking at all this a little more carefully, it seems that the module
> needs to be adjusted:
> 
> - "userinfo" needs to be optional, because an *absent* userinfo is
>  different from an *empty* userinfo.  (The latter has an "@".)
> 
> - ditto for "query"
> 
> - ditto for "fragment"
> 
> - "authority" needs to be optional within "uri"


To these points, I believe this is the case already.  Specifically:
for the nodes "userinfo", "query", and "fragment", which are all of type "string":
the node is optional, as the "mandatory false" is the default.
when the node is configured, the string may be empty (i.e., ""), because no "length" statement is specified and the default minimum "length" is zero (0)
the "authority" is optional, because the "presence" statement is specified.


> With those fixes, it seems to me that the module matches RFC 3986.  But
> maybe I'm wrong and someone can give an example that proves it.

I'm hopeful that no changes are needed to the YANG.


> Dale

Kent