Re: Fallback to UTF-8

On Apr 24, 2008, at 23:04 , Jukka K. Korpela wrote:

> Henri Sivonen wrote:
>> Considering the real Web content, it is better to pick Windows-1252
>> than a hypothetical generic encoding.
> No, it's not because _in validation_ you don't need to make any  
> guess on
> the meanings of octets > 127 decimal., for example, checks for bad byte sequences in the  
encoding (subject to decoder bugs), looks for the last two non- 
character code points on each plane and looks for PUA characters.

> You're not supposed to render them
> (apart from echoing them along with error messages, but they're not
> markup-significant) or to process them in any way but treating them as
> data characters.

Rendering source extracts is a significant part of validator UI.

> If you assume windows-1252, then many possible octets will be  
> unassigned
> and you may well have the problem of having guessed something and then
> detected the guess must be wrong. The document could be in some other
> 8-bit encoding, or in UTF-8, or something else, and if you hadn't  
> bet on
> windows-1252, you would have analyzed the markup properly.

Right, but if non-declared non-ASCII is an error, the pass/fail  
outcome will be right even if for the wrong reason.

Henri Sivonen

Received on Friday, 25 April 2008 07:17:11 UTC