Re: [Uri-review] Request for review

[Timothy Mcsweeney <[e-mail]>]

Hi Larry, 

Just so its crystal clear in my head, you're saying that the original intent of 3986

was that the colon, and only the colon, was to be used as the gen-delim between

scheme and path regardless of implementation specific sytanx?  I have that right?

On May 31, 2020 at 5:16 PM Larry Masinter <[e-mail]> wrote:

• I noticed two of the original authors are on the email list.  It would
• be nice to hear what the original interpretation was.

I wasn’t going to respond, but if you want my opinion, it’s that

we should change the registration form for new URI schemes to make this more explicit,

And we should consider updates to BCP 35/BCP 115 (Guidelines and Registration Procedures for URI schemes to figure out a way to terminate this kind of discussion.

I note that this re-(mis-)use of fragment identifier isn’t so different from the URN wish to repurpose “#”.

That we never required media type registrations to explain their interpretation of fragment identifiers. We should.

I note that for HTTP we can say “the fragment identifier isn’t sent to the server, only the base full URL” but that is not actually required and isn’t true for “file” URLs since in that case there is no particular “server”.

I point out https://developer.mozilla.org/en-US/docs/Web/API/Navigator/registerProtocolHandler" rel="nofollow">https://developer.mozilla.org/en-US/docs/Web/API/Navigator/registerProtocolHandler

      The https://developer.mozilla.org/en-US/docs/Web/API/Navigator" rel="nofollow">Navigator method registerProtocolHandler() lets web sites register their ability to open or handle particular URL schemes (aka protocols).

     For example, this API lets webmail sites open mailto: URLs, or VoIP sites open tel: URLs.

Which allows you to define “drop” URLs (which of course start with “drop:”)

From: Uri-review <[e-mail]> On Behalf Of Timothy Mcsweeney
Sent: Sunday, May 31, 2020 12:16 PM
To: Michael Wojcik <[e-mail]>; [e-mail]
Subject: Re: [Uri-review] Request for review

>The issue here is that you're misinterpreting RFC 3986.

I noticed two of the original authors are on the email list.  It would

be nice to hear what the original interpretation was.

>To be honest, I don't understand why you're being so difficult about this.

Having a different perspective is no being difficult. 

Imagine the first color blind person telling everyone the grass is dark gray.

On May 30, 2020 at 10:01 AM Michael Wojcik < [e-mail]> wrote:

From: Uri-review [mailto: [e-mail]] On Behalf Of Timothy Mcsweeney

Sent: Saturday, May 30, 2020 06:20

To: Graham Klyne; [e-mail]

And if people want to make parsers that don't work with the spec it doesn't

become the spec's problem.

That's not the issue here. The issue here is that you're misinterpreting RFC 3986.

3986 section 3 is not ambiguous. The first production is:

URI = scheme ":" hier-part [ "?" query ] [ "#" fragment ]

The colon is explicit and not optional. A minimal URI consists of a scheme, a colon, and a hier-part. There's no wiggle room there, and no amount of casuistry regarding other parts of 3986 will change that.

Someone could also point to 1.2.3, where the language clearly notes that the colon is the scheme delimiter; or 3.5, which makes it clear that the hash symbol is always the fragment delimiter. But those arguments are redundant in light of the generic-URI top-level production that begins section 3.

To be honest, I don't understand why you're being so difficult about this. What's your motive for trying to find grounds in 3986 for repurposing the fragment identifier?

--

Michael Wojcik