W3C home > Mailing lists > Public > ietf-http-wg@w3.org > April to June 2010

RE: Explicit instructions on use of fragment in request URI

From: Eran Hammer-Lahav <eran@hueniverse.com>
Date: Wed, 21 Apr 2010 23:49:09 -0700
To: Julian Reschke <julian.reschke@gmx.de>
CC: "HTTP Working Group (ietf-http-wg@w3.org)" <ietf-http-wg@w3.org>
Message-ID: <90C41DD21FB7C64BB94121FBBC2E723438E5C7FAEF@P3PW5EX1MB01.EX1.SECURESERVER.NET>
I'm not looking for a MUST NOT. But it would be nice for the spec to say "btw, you can't use fragments here".

I always knew fragments are not allowed, but when asked, I couldn't show where that's defined. And because HTTP doesn't make it easy, OAuth has to make developers aware of that.

EHL

> -----Original Message-----
> From: Julian Reschke [mailto:julian.reschke@gmx.de]
> Sent: Wednesday, April 21, 2010 11:38 PM
> To: Eran Hammer-Lahav
> Cc: HTTP Working Group (ietf-http-wg@w3.org)
> Subject: Re: Explicit instructions on use of fragment in request URI
> 
> On 22.04.2010 08:19, Eran Hammer-Lahav wrote:
> > This came up in the OAuth WG. One of the flows used to obtain a token
> relies on the fact that browsers don't sent the fragment over to the server
> and uses it to encoded credentials visible only to the browser (server
> provides them via a redirect Location header). I was asked what stops the
> browser from sending the fragment.
> >
> > I spent some time trying to find where 2616 forbids including the fragment
> and the best I came up with is from 3986:
> >
> >     the fragment identifier is separated
> >     from the rest of the URI prior to a dereference, and thus the
> >     identifying information within the fragment itself is dereferenced
> >     solely by the user agent, regardless of the URI scheme.
> >
> > Mark pointed me to the definition of request-URI which is abs_path or
> absoluteURI from 2396, which in turn do not allow a fragment.
> >
> > Would it be possible to make this easier?
> >
> > Something like "the request URI MUST NOT include a fragment
> > component"... :-)
> 
> Not convinced.
> 
> a) Has this been a problem somewhere?
> 
> b) Making a BCP14 requirement on something the syntax doesn't allow in the
> first place doesn't feel right to me. If the request contains a fragment
> component, it's invalid. Period.
> 
> Best regards, Julian
Received on Thursday, 22 April 2010 06:49:54 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:51:18 GMT