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

[Kent Watsen <[e-mail]>]

Hi Martin,


> On Dec 3, 2025, at 2:27 PM, Martin Thomson <[e-mail]> wrote:
> 
> Hi Kent,
> 
> On Thu, Dec 4, 2025, at 06:03, Kent Watsen wrote:
>>> URIs are more complicated than your decomposed structure allows for.  That's the problem.  If you are going to represent a URI, it really has to be a string.
>> 
>> I believe you, but I'm unsure what I missed in RFC 3986...
>> 
>> Does this regard the percent-encoded form of URIs that don't follow 
>> the "normal circumstance" mentioned in the first paragraph here: 
>> https://datatracker.ietf.org/doc/html/rfc3986#section-2.4?   Would it 
>> help for this draft to state that the values MAY be percent-encoded?
> 
> I was inclined to ask what the "ietf-inet:uri" definition says, but it really doesn't say anything.  Which is OK.

Sure, same as with the "ietf-uri:uri" definition.

But I'm still trying to square the statement "URIs are more complicated than your decomposed structure allows for".  So far, they still appear to be 1-1 to me.  Can an example where conversion doesn't hold be provided?  A pointer to some text in an RFC would also be great.



> That section says this:
> 
>> Once produced, a URI is always in its percent-encoded form.
> 
> In other words, don't worry about that, because octets that are not in the URI grammar will be percent-encoded by the thing that produces the URI.  (If that's you, great. You get to learn how to make a URI. No doubt you will get it "wrong" by some objective measure, but that's OK, because everyone does. The way we cope is that most URI-handling software will either manage or not work.  So you can test.  That's not a great story, but it's the one we've got.)

I'm familiar with percent-encoding.  For what it's worth, I'm also author on RFC 8040 (RESTCONF) and have personally developed a RESTCONF server capable of encoding URI containing query parameters into URI query parameters, which likely maximizes the need for percent-encoding!  Of course, I used a library to do the work, specifically the "quote" and "unquote" routines from urllib.parse.  In any case, the code was right on the second commit ;)


Kent // author of draft-ietf-netconf-http-client-server