W3C home > Mailing lists > Public > w3c-wai-ig@w3.org > July to September 2006

Re: correcting URIs - was User friendly 404s reconsidered

From: Charles McCathieNevile <chaals@opera.com>
Date: Thu, 20 Jul 2006 13:03:06 +0200
To: "Patrick H. Lauke" <redux@splintered.co.uk>, "Jesper Tverskov" <jesper@tverskov.dk>
Cc: w3c-wai-ig@w3.org
Message-ID: <op.tczibg12wxe0ny@62-151-100-1.jazzfree.ya.com>

On Thu, 20 Jul 2006 12:36:57 +0200, Patrick H. Lauke  
<redux@splintered.co.uk> wrote:

> Quoting Jesper Tverskov <jesper@tverskov.dk>:
>> If we add /vista an error message is not returned but the URL is
>> automatically corrected to /windowsvista.
>> This is misuse of the error message turning it into some sort of search
>> engine not even notifying the user.

> So, just to recap: 404s should never attempt to help the user actually  
> achieve their goal?

I think this is a misinterpretation of what Jesper says...

> If an error can reasonably be corrected, the application should not  
> attempt to do that, but rather give a STBU message?
> I'd say that when there is an *obvious* way to correct an error,  
> usability would suggest that it's a good thing to actually do so. How do  
> you feel about Apache's mod_speling, for instance?

mod_speling is a tricky thing to use. Once upon a time W3C turned it on  
and used it extensively. It turned out to cause a lot of problems and was  
rapidly turned "mostly off" (it is applied to regularise capitalisation I  

The basic problem is that it is effectively minting a whole swag of URIs,  
and they are not necessarily consistent (i.e. don't always point to the  
same thing, a violation of the "cool URIs don't change" idea that I think  
is pretty fundamental to the real power of the web).

> And, to repeat what I wrote over on the Accessifyforum:
> if you see users typing an outdated/wrong URL (or, stretching it even  
> further, following a broken link i.e. feed erroneous/outdated  
> information into the system that is your web server) as an "input  
> error", then providing a list of possible matches or any other  
> information that may help the user is actually recommended...

There is a crucial difference between providing information, and selecting  
among that information without asking the user. In some situtations it's  
fine, in others it is actually detrimental in the long term. In China, I  
am told, it is common to use the fact that Firefox defaults to an "I'm  
feeling lucky" search from google if you enter something that is not a  
legal URI as a way to find companies whose names are written in chinese.  
For some proportion of well-known companies, organisations, schools,  
people, etc this is fine. But the reliance on Google (or any search  
engine) not just for information but to make a further choice based on  
that information is a substantial gamble, and in the long term there is a  
risk that it significantly restricts what people find. There is a valid  
question about whether that matters in practice, of course.

I am inclined to go along with Jesper on the question of auto-correcting  
for things not found, in light of my experience. A list of possibilities  
is fine. Randomly papering over any appreciable amount of the infinitely  
many gaps in the URI-space, in ways that are not consistent, strikes me as  
a bad idea.



   Charles McCathieNevile, Opera Software: Standards Group
   hablo español  -  je parle français  -  jeg lærer norsk
chaals@opera.com          Try Opera 9 now! http://opera.com
Received on Thursday, 20 July 2006 11:03:22 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 20:36:28 UTC