1. When Google first unveiled video.google.com, you had to use their patched version of VLC to play videos there, and if you used Google's patched version of VLC, it would only play items hosted on http://video.google.com. (They have since changed over to a Flash player using FLV, which is not like this).
2. If you want to watch news video on MSNBC.com, you have to be using Internet Explorer or you will get an error message saying that IE is required. It's not good enough to have Windows Media Player as either a standalone or plugin, and it's not good enough to use an alternate media player like VLC which is capable of rendering Windows Media. It's possible that there are technical issues associated with this (probably the site relies on features specific to IE) but not that these technical issues were insurmountable.
3. If you want to browse the iTunes music store over the web, you have to be using iTunes as your web browser. With Internet Explorer or any other browser aside from iTunes, you get an error message saying that you must use iTunes.

So here's a potential litmus test: what makes media open is whether any potential pair of clients and servers can work together to fetch it and render it.

If the protocol is an open standard like HTTP we know in advance that any server can return a given file and any client can fetch it, but this isn't good enough -- all three examples above use HTTP. If the media is in a file format for which there exist multiple players we also know in advance that any client (well, most) can render it, but this isn't good enough -- all the examples above use formats for which there are multiple players.

When we talk about open media technology, the subject is usually file formats which are proprietary to the maker of the player software. That's a Bad Thing, and it's certainly not open, but it's not what I'm talking about here. One of the major factors blocking media on the web is that it is often restricted so that network effects are impossible. In my examples there are three clients and three sites, and if the whole internet followed this approach there would be a different piece of software for every possible site.

Flash is sometimes another flavor of this general problem. If a site provides their own Flash audio/video player, restricts it to loading content only from the same site, and restricts content to only being loaded by that player, there is effectively a different client for every host. That is what happens with SWF files which bundle the player logic and content together.

A fatal flaw in my litmus test is that it excludes the recommenders. Even if there did exist a healthy ecosystem where any client and server could pair up, it wouldn't do much good if clients had to go through centralized sources to find new content, which is how it works in meatspace.

Clients are forced to do exactly that when link publishers are required to host the media they're linking to. On the plain old browser web that's a silly idea which nobody takes seriously, but with media files direct links are still a problem.

First off, there are many people who think that direct linking is stealing bandwidth.

Second, many people are confused about the distinction between linking and copying. They feel that copyright grants a veto over who can link. For example, there was a prominent videoblogger who threatened to sue a sleazebag site which linked to his video, and many members of that community felt that he was in the right. Similarly, I can't imagine that a rights holder could be sued for linking to an unauthorized host of his own material, but I can easily imagine him suing somebody else for the exact same link.

Lastly, there are technical problems caused by badly conceived protocols and software. For example, if I publish a URL in an RSS enclosure I have to set the media type as an attribute (it's required), but if I am not the host of the URL I have no control over the media type. Any client and any server can get together, however third parties can't publish links without violating the spec.

Given that, my litmus test for open media is whether anybody fetch, render or link -- it's a three-way system.