Re: "image analysis heuristics" (ISSUE-66)

On Sat, 6 Feb 2010, Lachlan Hunt wrote:
> 
> Based on the discussion, I've come up with the following proposed criteria:
> 
> 1. Permits user agents to apply any repair technique

Agreed.


> 2. Does not list specific techniques.

I think we do a disservice to users by not mentioning explicit techniques.


> 3. Informatively references the techiques described in UAAG (either 1.0
>    or the 2.0 draft)

We already have a reference to UAAG in the "Recommended reading" section. 
I don't think we should have fine-grained cross-references throughout the 
spec, because that way lies a maintenance nightmare, and it's not actually 
that useful (UA implementors should read the whole UAAG).


> 4. The text is located in a section dedicated to describing
>    implementation requirements and is clearly distinguished from
>    authoring requirements.

Agreed.


> 5. Does not imply the use of futuristic technologies.

Image recogition is not a futuristic technology, and nor is OCR.


> 6. Does not imply that any technique can reliably determine author
>    intent.

Agreed.


> 7. Indicates that it is about providing additional information about the
>    image, which may help the user to understand the image's content or
>    purpose.

It's about aiding the user's comprehension of the page; if that can be 
done without the user ever being informed of the image, then I don't see 
why that should be non-conforming. So I disagree with this.


> In my own view, the current text [...] partially fails #5 by mentioning 
> "heuristics" without clearly describing what would or would not be 
> classified as such

How can one allow any repair technique while enumerating all the 
techniques? I don't think those two requirements are compatible.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Saturday, 6 February 2010 11:03:26 UTC