Brion Vibber kirjoitti:
The simplest possible interface here would be to provide an HTML fragment containing the challenge; though that brings with it potential issues of safety/sanitization or HTML rendering requirements.
No HTML in API responses, please. If I wanted to parse HTML, I'd be screen scraping the wiki interface anyway, not using the API.
URL + MIME type, plus possibly some explanatory text, should really be enough for any CAPTCHA. They're supposed to be pretty self-explanatory anyway.
One could even use "data:" URLs for short in-line CAPTHAs if one wanted.