463955 - Planet Mozilla should strip autoplay attribute from <video> element

Status: RESOLVED FIXED

(Websites :: planet.mozilla.org, defect)

Description

[Jesse Ruderman]

http://www.rumblingedge.com/2008/11/09/mark-surman-interview-in-sept-08/ contains a video that starts playing right away due to having the autoplay attribute.  This means loading Planet Mozilla will make noise until it falls off the bottom :(  I think Planet should strip the autoplay attribute, so the video only starts if a reader wants to to play.

Comment 1

[Asa Dotzler [:asa]]

yeah. I agree. I don't know how to do it, but I agree that we should.

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

This might help:

http://www.0xdeadbeef.com/weblog/?p=1127

If I'm reading Chris Blizzard's post correctly, it's feedparser in Planet software that does the stripping. Maybe contact Sam Ruby or someone else?

Comment 3

[Jeff Walden [:Waldo]]

http://svn.mozilla.org/projects/planet/trunk/planet/vendor/feedparser.py

Remove autoplay from the array that includes it as a string and that would probably do it; that said, you really want to add the controls attribute at the same time.  Also, this is a bit hackish in allowing autoplay on any element, not just video/audio, but that seems to be a more fundamental problem that I wouldn't suggest tackling without doing it upstream.

Comment 4

[Reed Loden [:reed]]

Sam, can you fix feedparser upstream for this, as I bet others feel the same way.

Comment 5

[Christopher Blizzard (:blizzard)]

Here are my thoughts on this, welcome or unwelcome. :)

1. It's a valid attribute and is part of the spec.  It's not dangerous (just annoying) so I would hate to see feedparser pull it out.  (Information loss that early in a pipeline is always bad.)

2. User agents should solve this problem.  I'll start a thread on platform and firefox about it.

3. Planet (not feedparser) should probably make it an option to pull it.  Mostly because if you're pulling a huge page it's hard to find the video that's playing.

I don't know if the planet/venus software can do #3 or not.  But feedparser seems like the wrong place to make that change.

Thoughts?

Comment 6

[Sam Ruby]

I've made the change: http://code.google.com/p/feedparser/source/detail?r=294

A case can be made that feed parsers shouldn't do sanitization, that should be done at a separate layer.  But that's not the current approach:

http://diveintomark.org/archives/2003/06/12/how_to_consume_rss_safely
http://diveintomark.org/archives/2002/10/28/we_have_arrived

I personally agree that sanitization should be done separately, controlled by planet config options, and since refactoring planet software is something I seem to do as a hobby, I started such a refactoring over a year ago

http://intertwingly.net/blog/2007/12/19/Yet-Another-Planet-Refactoring

What is slowing me down for the moment is that I really want to make this HTML5 and DOM based through and trough.  What I would love to do first is to take hsivonen's parser and make it spit out C/C++ code which can be directly called by a language like Python or Ruby.  Unfortunately, the current code output is very mozilla specific, and I haven't had the cycles to get back to it.

Comment 7

[Reed Loden [:reed]]

I'll pull this to Planet Mozilla in an hour or so.

Comment 8

[Robert Accettura [:raccettura]]

Not sure if this should block or depend (i guess it's up for interpretation).  Regardless it's related.

Comment 9

[Reed Loden [:reed]]

I did this the other day.