[Commons-l] We should permit Flash video playback

[Gregory Maxwell]

On 7/20/07, Erik Moeller <erik at wikimedia.org> wrote:
> On 7/20/07, Gregory Maxwell <gmaxwell at gmail.com> wrote:
> > By doing so we lose the ability to offer a complete record of our
> > content, and and the ability to offer easy 'one stop' duplication of
> > our collection for backup and other purposes.
>
> http://download.wikimedia.org/
>
> Image tarballs
>     There are currently no image dumps available. Check back in mid-2007.


Yes indeed. In fact I understand that our own internal backup
situation for images isn't much better than the dumps...

But by your argument we should shut down enwiki and hand it off to
archive.org: After all, our dumps are highly unreliable: there is a
new partial set one from the 16th, but before then and for some time
back there was no dump at all.

> We don't seem to be doing very well on this front one way or another.
> The Internet Archive _specializes_ on redundant, fail-safe archiving.
> I'm sure providing easy ways to access all uploaded videos would be
> much easier for them than it is for us.

We need redundant fail safe storage etc for our images as well. We're
failing. The solution is to stop failing, not hand over the ropes to
someone else.

> > A lot of the content that Archive.org hosts would be deleted on
> > commons for copyright reasons. Historically they have handled
> > copyright by exception rather than proactively.
>
> It's an _archive_. Their point is to slurp up as much material as
> possible. It's wonderful that an organization with this purpose exists
> which is willing to push the limits of what is permissible under
> copyright law, and if we only push free content to them, rather than
> adopting their criteria for inclusion, their mission in no way
> denigrates ours. We could selectively whitelist some of their
> collections as being acceptable for use within our projects.

What they do is their business and I wish them luck.
At the same time, we're here to create content which provides freedom,
not get into pissing matches with copyright holders.

> > Yet some members of our development staff do not respond to emails
> > about video playback support in Mediawiki.
>
> By "some members of our development staff", do you mean Tim or Brion?
> ;-) Let's face it, we don't currently have the staff support to go
> much beyond just keeping the sites running.

Tim.

> > Where have you been?  :) We now have automatic in browser playback
> > that works for a majority of readers without downloading anything
> > additional.
>
> We still don't support embedding video directly into articles (the
> Archive has a nice implementation where the player loads in the brower
> once you click the preview).

We could have in about 15 minutes- 20 minutes, I'm not kidding. I'm
not a fan of inline players because they mess up layout. No one has
asked me for it so I've avoided doing it.

>What are you basing the assertion
> "majority of readers" on?

Our popup "(>) play in browser" system loads a launch page off a box
with an access log.  Due to java security it also fetches the video
through a proxy on the same system.  We can tell if the player worked
for an IP if the same IP both fetches the launch page and completes a
media file.

Previously it worked for 81-82% of the people who clicked the play button.
I pulled yesterdays numbers and it was only 72%. Both are still
clearly a majority. :) This is in the same general range as flash
penetration, at least for flash 8+... and  it's way better than before
I setup the player: about 0% could just click a theora file and play
it ;)

I made some changes to the player autodectection code a few weeks ago,
so I suspect I botched detection for some users, resulting in the
decline rather than an actual change in browser abilities. If thats
the case it's entirely my fault: I put out the last set of detection
changes without much testing. :(

Obviously the viewer numbers are a bit self-selecting: If it didn't
work before people are less likely to try it later. But even on day
one I had success levels well over 50% and it works for a lot more
people now. Also, much of our traffic is drive by, which should reduce
the memory effect.

> > Jeez Erik. With the exception of flash we already have that.
> > *Including* HTML5 <video/> support, which works in Opera.
>
> It does? I'm using Opera 9.2, and it only shows me the Java player.

You need the 9.5 beta, which I think is still windows only. It works
fine in Wine however.

> Which does work, though it initially showed me a tiny version of the
> video (and the full video on reload). My experience with Java applets
> has generally been very negative, with memory usage and initialization
> time often being prohibitively high, and debugging for different
> platforms being very difficult.

Java is strongly disliked in most of the Linux using crowd, the
overwhelming majority of Linux users don't have it installed. For them
the VLC, and Application/Ogg support should take up the slack. (Both
are preferred by the player over Java if they appear to be available)

I am not a Java fan. If you know some java fans, please send them to
me so I can pawn off any Java coding that I need.  That said, it's a
norm in the Windows using world, especially for folks more on the
bussiness apps and academics than the web games side of the browsing
spectrum.

> We should carefully look at the experience of the vast majority of
> users who are a) on Windows, b) use Internet Explorer or the default
> install of Firefox. I would also bet that most of them don't have Java
> installed, but it seems hard to get numbers on that. Being able to
> play video directly in the article being viewed is also highly
> desirable.

Most of the people who hit play have Java. 84% of the plays come from
Windows users.  72% are not using Firefox.

We can have exact numbers for Java support for anyone who hits 'play',
but I don't have them now because of caching and Java being the least
preferred. At the very minimum it is over 50% (since half the IPs are
loading the JAR for the player).

Right now I believe the player still tends to work better for
joe-average Windows user than joe average techno-dork, but thats only
an educated guess.

It needs to improve, but to claim that we're causing a huge hassle for
the sake of supporting free formats really isn't supported by what I'm
seeing.


> > Wikimedia is in the US. How do you propose we pay the licensing fees
> > for the codec patents we are using, should we be presented with a
> > bill?
>
> I think we should simply ask the current patent holders whether they
> would grant us non-commercial rights to use the relevant codecs. Yes,
> such rights won't trickle down to third parties, but we would offer
> the Theora files and promote Theora for this reason. The right could
> be time-limited, and renegotiated regularly.

So will we also be asking for a free copy of MS [[SharePoint]] so we
finally have WYSIWYG editing by allowing people to edit pages in
Microsoft .doc format? :)  Not a serious question of course, but one
should ask where the line is with your suggested thinking.

I don't think we should distribute our content using any licenses or
technology which we can't make available to others. Thats been our
legacy, and there are strong arguments for it.

Doing so is a violation of the spirit of any copyleft license, since
you've encumbered a copy of the work, if not the actual terms. I know
you're not a fan of copyleft, but it is what it is.

> > If someone doesn't step forward and push for unencoumbered formats
>
> I'm 100% in support of "pushing unencumbered formats", but not at the
> expense of usability for the majority of users.

Majority? If thats the criteria then we already passed it when I
created  the in browser player.

I'd like to see it work better and for more people... and there is a
lot of development still going on. (For example, I will very soon have
the Java mode displaying video at the right sizes).

I don't personally agree that there is any percentage of users who
can't view it that makes it okay to compromise freedom: If people
can't view it we should spend effort making it work without the
compromise, making the world a place where people can freely share
knowledge without format taxes.

...As it turns out, however, what we have is working for a lot of
people. So thats an argument we shouldn't have to have.

> > We already have a solution that works for a lot of Windows users
> > without installing anything most don't already have. (Flash isn't
> > installed by default in Windows either.. but most have it. Java
> > penetration isn't quite as deep, but for our readers it does
> > remarkably well)
>
> I'd love to see hard data on this. Failing any already existing data,
> a large scale video survey on our projects could be a good first step.

All you had to do was ask... :) I don't have more than two days of
logs for the bounce page loads, but I do have them going back to March
for successful media plays.

Yesterday 24313 out of 33907 distinct IPs that pressed the play button
completed playing a media file with the in-browser player. (this
traffic level is historically low but there is a lot of day to day
variance)

Since march 1,519,989 distinct IPs have successfully played a media
file with the in browser player I operate. The total number of files
played in that span is 4,103,018. (I can't say attempts over that
span, alas, due to a lack of logs).

Is there room for improvement? Hell yes. But thus far I've been going
at this alone.

Do we need to compromise freedom to improve things? No.

Do we need to compromise freedom to support lots of people? We already
do without any compromise in freedom.

Cheers.