Re: Robin's alt attribute report

On 24/11/2012 14:49 , Marcos Caceres wrote:
> Robin, can you share some thoughts about what you were trying to
> achieve ?

I mostly wanted to know if the alt usage advice given in the HTML 
specification was paving the cow paths or not. It looks like it wasn't.

> It's kinda unique in the that you needed to look at the use
> of an attribute in the context of surrounding text. Also, I noticed
> that there is no actual analysis of the data in the report … I wonder
> if that can actually be automated in some way. At least, to get a
> statistical view of for when alt is used, what is the mode, mean,
> median, and standard deviation for each site, etc.

Statistical information is only useful if you know how to read it 
properly, which is the case of only few people.

In this instance anyway the final step requires human evaluation. The 
report is just making the information available in context. So there 
isn't anything to apply stats to.

Thais said, one thing is clear from processing this data set: anything 
that requires more than grep is slow. It's not a huge deal in that you 
can just let it run (which is what I did) but it's still annoying.

On Twitter a bunch of us hashed out potential solutions to that. The 
best bet is to use an XML database (yes) and inject the parsed HTML DOM 
into it with minimal discrepancies (there will be some, they can be 
documented). The results then becomes instantly queryable, blazing fast. 
It's not overly complicated, and it would be useful to expose that as a 
public service.

-- 
Robin Berjon - http://berjon.com/ - @robinberjon

Received on Monday, 26 November 2012 09:34:29 UTC