W3C home > Mailing lists > Public > public-html-comments@w3.org > July 2008

Re: 4.13: URI decomposition - non-standard terminology

From: Frank Ellermann <nobody@xyzzy.claranet.de>
Date: Sat, 12 Jul 2008 08:48:32 +0200
To: public-html-comments@w3.org
Message-ID: <g59k2t$1u8$1@ger.gmane.org>

Ian Hickson wrote:
>> I did
> Ok, well, whenever you get around to it please do let me know
> if there are any issues that need resolving in that section.

When I wrote "I did" I meant that I found 4.13 in the "public"
draft.  That talks about "protocol" as the OP said, this is no
issue as you said, but the OP said "for example".

Whatever he had in mind, IFF it was <hostport>, then yes, this
is marked as "obsolete terminology" in appendix D.2 of STD 66,
your expansion host [":" port] is precisely what it used to be,
but you could as well avoid to talk about <hostport>.  

I hope that's clearer now.  You've already mentioned the issue
with using "URL" for something that is no URL as far as STD 66
and dozens of RFCs are concerned (some of them not yet updated
to reference STD 66, ditto various popular document formats).

> The default port numbers are all here:
>    http://www.iana.org/assignments/port-numbers

Not really, these are reserved ports per protocol, this doesn't
tell you which URI schemes are associated with which protocol.
E.g., "news:" is no protocol, and "nntps" is no URI scheme, but
"nntp" is a protocol and an URI scheme, and has a reserved port.

> it isn't clear what browsers should do with URLs of schemes
> they don't know (say, x-foo-bar://...). I should test that
> and make sure it's covered.

ACK, I'd love to know how I can tell BBFH that xmpp: is a good
scheme, and that I have a jabber client.  Convincing BBFH that
gopher port 43 (among others) is perfectly fine would be also
nice, or at least better than inventing any whois: URI scheme.

> Could you give a code example that illustrates the case you're
> concerned about?

<http://example.com:> and <http://example.com:/> are allowed in
the STD 66 syntax.  BBFH ignores the colon, apparently, I've no
tool to see what it actually does at the moment.  I'm not sure
what it does with <http://example.com:0x50>, but apparently it
tries example.com:50.  New test with <http://example.com:0x80>,
I get example.com:80.  BBFH is awful.  Your draft says that it
should try example.com:0  But I don't see the point, why not
just say "invalid URI" and give up, if the port is nonsense ?

> What is BBFH?

Bastard browser from hell, also known as Firefox  
It supports <http://[]/> (not a proper URL)
and <http://[::ffff:]/> (a valid URL), but
also <http://0xd0.0x4d.0xbc.0xa6/> (again no proper URL).

> Which version of the spec are you reading that it doesn't 
> have a section 5.9.8?

The first link in <http://www.w3.org/html/wg/html5/> is the
latest published version <http://www.w3.org/TR/html5/>, and
that has no section 5.9.8, chapter 5 ends with 5.6.  But it
works with BBFH. 

Received on Saturday, 12 July 2008 06:48:45 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:26:25 UTC