Re: [apps-discuss] Fwd: FW: New Version Notification for draft-kerwin-file-scheme-13.txt

[Sam Ruby <[e-mail]>]

On 01/04/2015 07:22 AM, Matthew Kerwin wrote:
>
> On 2 January 2015 at 00:21, Sam Ruby <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>     A few concrete examples are here:
>
>     http://www.ietf.org/mail-__archive/web/apps-discuss/__current/msg13511.html
>     <http://www.ietf.org/mail-archive/web/apps-discuss/current/msg13511.html>
>
>     It is my hope that by comparing notes we can converge.
>
> ​To be honest, a big reason for using RFC 3986 as a normative reference
> is so I don't have to deal with things like hostnames. If those
> definitions are good enough for the shiny new HTTP/1.1 (which actually
> uses hostnames) then surely they're also sufficient for file: (which
> often doesn't).
>
> BTW who on earth writes localhost as "2130706433"? Or more
> significantly, who on earth *accepts* that as a valid URL?

I cited three specific examples in:

http://www.ietf.org/mail-archive/web/apps-discuss/current/msg13511.html

All three of the below are accepted by Chrome, and possibly others:

file://2130706433/foo
file://[2001::1]/foo
file://０Ｘｃ０．０２５０．０１/foo

Note that RFC 3986 requires square brackets around IPv6 names, but your 
draft does not.  The final example involves IDNA processing. 
Unfortunately, IDNA processing can't be avoided, nor does RFC 3986 
specify it.

>         I guess I could add support for backslashes. I suspect that
>         would end up
>         in a non-normative appendix, along with the "|" drive letter
>         separator
>         (as suggested in an earlier message, and currently being worked
>         into the
>         next draft.)
>
>
>     I don't understand the thought process here. The current
>     Internet-Draft takes care to document a select few browser specific
>     syntaxes, but many others (including the ones you mention above) are
>     not included.  What is the selection criteria you are using?
>
> For the most part: things that aren't silly (big-endian integer encoding
> of IPv4 addresses), things that aren't widely touted as obsolete (like
> backslashes[1]), and as much as possible, things that go against the
> parent standards (like trying to jimmy "/c|/" into RFC 3986).
>
> Importantly, I don't want to include every codepath in every
> implementation that exists, but I don't want the definition I write to
> *not* work in any implementations.
>
> [1] http://blogs.msdn.com/b/ie/archive/2006/12/06/file-uris-in-windows.aspx

You have a definition for a "windows-path" and a "unc-path".  These are 
unlikely to work on Android or OS/X.

Firefox does not appear to support host for file: URIs.

> --
>    Matthew Kerwin
> http://matthew.kerwin.net.au/

- Sam Ruby