ChromeDriver proposal using DevTools/Extensions

[Ken Kania]

ChromeDriver is an implementation of WebDriver, a w3c working draft that presents an API for out of process website testing.  This allows website devs to write tests that find elements on a page and manipulate them using the mouse and keyboard.

Here's the design doc for the new ChromeDriver: https://docs.google.com/document/d/1FAuPLN92rSinz55J7R3H1hHmmQBuXiO3SUGIeZW1lI0/edit

The main points of discussion I see are:
1) Whether to add new DevTools commands for user input or expose EventSender via special flag/setting

2) Whether to implement ChromeDriver as a JavaScript page hosted in Chrome or as a separate python, java, or C++ application.

3) The other proposed changes to DevTools

Awhile back, Zori and I sent out a proposal for revising the automation system for WebDriver.  This approach was favored over that because 1) chromium is trying to reduce dependence on the automation system, and 2) most of the stuff WebDriver needs to do is in the WebKit layer.

Feel free to send comments by email or on the doc.

Thanks

Ken

[Paweł Hajdan, Jr.]

On Fri, Sep 14, 2012 at 12:47 AM, Ken Kania <kka...@chromium.org> wrote:

Awhile back, Zori and I sent out a proposal for revising the automation system for WebDriver.  This approach was favored over that because 1) chromium is trying to reduce dependence on the automation system, and 2) most of the stuff WebDriver needs to do is in the WebKit layer.

Yeah, if you can design things so that you don't need automation, it sounds good to me. +1 for this DevTools-based approach (from my automation-POV). 

[Ken Kania]

I've updated the doc to incorporate some feedback.  Please take another look at the Architecture section and let me know what you think.

Ken

[Paweł Hajdan, Jr.]

I was not able to see the revision history of that Google Doc, but the Architecture section sounds fine.