webkit2png or similar for site previews

[olliencc]

So I proto-typed a solution yesterday which used an iframe to provide a preview of all the web sites. This basically sucked and resulted in git reset --hard HEAD^. Why did it suck?

• x-frame-options
• sites requesting auth
• sites doing iframe busting

All led to a very crappy user experience. The proper solution is use something like webkit2png (whilst trying to not get exploited using a crappy old version of webkit)

• https://github.com/AdamN/python-webkit2png/

This needs some thought to balance security / functionality.

[divemonkey]

Have a look at Selenium Web Driver which will let you use Chrome or Firefox:

http://coreygoldberg.blogspot.co.uk/2011/07/python-taking-browser-screenshots-with.html

[olliencc]

only concern I have with that approach is when it's run on a server it creates a massive dependency

[divemonkey]

Probably not much more than WebKit2PNG. The only concern would be speed and resources with Selenium. I am not sure what the overhead is for Webkit but Selenium is much higher.

How about outsourcing it completely to something like http://web-capture.net/ which would remove any client and server dependency?

[olliencc]

I believe this is a viable solution - http://www.chrisle.me/2013/08/running-headless-selenium-with-chrome/

[neonbunny]

There's a list of 3rd party screenshot services at http://snapcasa.com but all the good ones are commercial. I experimented with spawning phantomjs with a screenshot script for every www sub domain, but although it worked, the number of processes was very resource intensive (and 3 never terminated).
There's a list of more fully featured screenshot tools on phantomjs's related projects page (http://phantomjs.org/related-projects.html). I favour one like https://github.com/gre/screenshot-webservice which manages the resource instensity and could be deployed on another server/load balanced cluster for better security/performance if required.

[olliencc]

yeah I don't see this going to be an easy one.. risk of compromise is just massive as well due to crusty webkit versions under the hood.