Re: parsing URI (references) according to RFC 3986

On Sun, Jun 19, 2011 at 9:26 PM, Boris Zbarsky <bzbarsky@mit.edu> wrote:
> On 6/20/11 12:03 AM, Adam Barth wrote:
>> Would you be interested in treating \ and / equivalently in more
>> schemes in order to converge behavior with IE, Chrome, and Safari?
>
> I'm not the one making that call; in fact I'm trying to get out of the
> business of dealing with our networking code...  So I can only comment on
> what we do right now and maybe on why we do it.

Without the relevant decision makers, we're unlikely to make progress
on a spec that matches implementations.  One reasonable perspective is
that not having a reality-matching spec here isn't causing much pain
in the world, so we should abandon this project.

> Historically we have been unwilling to do the above (obviously, since we
> haven't done it); I can't recall how much was a matter of doing what the
> spec says as much as possible and how much was specific use cases that the
> behavior precludes (e.g. using unescaped \ in filenames when the file has
> that name on the (non-Windows) server).

Of course, folks who want to name such files will want to use the
escaped form of \ in order for their site to work in browsers other
than Firefox.

> Note that the Chrome and Safari behavior is limited to some particular URI
> schemes or classes of URI schemes too.  Testcase:
>
>  data:text/html,<b>bold<\b>or not
>
> which shows the bold text "bold<\b>or not".  I don't have IE9 on hand right
> this second to check its behavior.

Correct.  It's limited to "hierarchical" schemes, such as http, ftp,
etc (at least in Chrome).

Adam

Received on Monday, 20 June 2011 04:45:32 UTC