Re: Error handling in URIs

Tim Bray wrote:
> Also, I'm not enthusiastic about writing standards unless
> there's an obvious pain point that needs to be addressed.  If the
> implementors are in general doing the right thing in a compatible way,
> is any further spec work required?

Here's a real case:
Validator.nu has a feature called "Image Report". It reads a document  
from the Web, keeps a stack of the HTTP-level URI, <base> and xml:base  
context and resolves <img src> to absolute URIs according to IRI  
rules. Now if the input document is not encoded in UTF-8 and the the  
src attribute contains non-ASCII in the query string, the result no  
longer dereferences to the same image it would in a browser. (This is  
an untested statement based on the assumption that <img src> behaves  
like <a href>, which isn't a safe assumption.)

Thus, even if the incumbent browser vendors have figured this stuff  
out, everyone else who seeks to consume real Web content will  
initially write incompatible software if reality-based conversion to  
ASCII-only URI isn't written down somewhere.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Wednesday, 25 June 2008 10:17:18 UTC