W3C home > Mailing lists > Public > public-html-a11y@w3.org > March 2010

Re: keep CAPTCHA out of HTML5

From: Maciej Stachowiak <mjs@apple.com>
Date: Wed, 17 Mar 2010 15:56:03 -0700
Cc: 'Silvia Pfeiffer' <silviapfeiffer1@gmail.com>, "'Gregory J. Rosmaita'" <oedipus@hicom.net>, public-html-a11y@w3.org
Message-id: <886B3837-9857-4BFE-A6A4-1A16C7D48434@apple.com>
To: John Foliot <jfoliot@stanford.edu>

On Mar 17, 2010, at 1:58 PM, John Foliot wrote:

> Silvia Pfeiffer
>>
>> Is there no way at all to transfer a CAPTCHA onto a braille device?
>>
>> Just curious.
>
> Hi Silvia,
>
> No, as it then would also be accessible to bots, etc.  The whole  
> point is
> that CAPTCHAs (visual or audio) are actually testing for physical  
> ability
> rather than cognition - you can see something or hear something,  
> actions
> that machines can replicate but not accurately process - thus, the  
> logic
> goes, you are human and not machine.

What it's really asking for is cognition that only a human is believed  
capable of (including interpretation of visual or audio stimuli). Some  
captcha-like gates I have seen ask a topical question in plain  
english. I am not sure if these are as effective as the traditional  
kind.

>
> But if a disabled user requires a machine to interact with technology
> (i.e. Adaptive Technology), then the very task that CAPTCHAs have been
> created to do (frustrate machine processing) impacts the user  
> *reliant* on
> machine processing. (And we won't get into the whole issue of some
> people's belief in OCR to save the world, or speech-to-text for
> captioning, because if these technologies did actually work with 100%
> accuracy, they would completely negate the 'security' around CAPTCHAs)

Actually, the reason CAPTCHAs are so ridiculous is because the state  
of the art on OCR is so good. It's very hard to make something that a  
human can read but a machine can't. CAPTCHAs that are currently in use  
are barely legible even to people of normal vision.

Regards,
Maciej
Received on Wednesday, 17 March 2010 22:56:37 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 04:42:06 GMT