Train of Thought

I’m now about half way through my sabbatical year, so I thought a progress update would be in order. You may recall that I wanted to see if I could do two things:

• Integrate more cycling into my habits.
• Find out if I have the capacity to start my own software company.

Taking the cycling first, I haven’t much. Indeed, I cycle as infrequently as when I had a full time job. An early start has fizzled out, and my main activity has been to watch this year’s Tour de France. However, I got a nice new cycle computer for my birthday last week, so perhaps the trick of ‘new equipment’ will get me going (it has worked in the past).

On the software company front, I have made more progress, although I don’t claim to have a going concern yet! I have formed Transmission Begins, and developed its first product, My Own TV Channel. This is now just starting to be used by people other than me. It’s well received whenever I discuss it, so my current plan is to improve it somewhat and then focus on building a service business around it. The software itself is available for download with an open-source licence.

Apologies to those of you who don’t know a speaker of a little known language, ook. Clearly there is a sufferer in Grantham.

No pixels were photoshopped in the creation of this post.

Inspired by my encounter with the Miro team, and mindful of a need to broaden my skillset beyond client software, I’ve been busy the last few weeks. One comment that was made at the Miro presentation was that most of the channels are hosted over plain http, and make no use of the bittorrent client in Miro. In the pub discussion afterward, we seemed to conclude that bittorrent server software was still too hard to use.

Curious, I started looking around. Most of the bittorrent software I could find was limited in some awkward way - perhaps implemented in a less popular language, or designed to handle the needs of the high traffic pirate content servers rather than small providers of their own content. Two services stood out: Amazon’s S3 service lets you turn any hosted file into a torrent by adding the ‘.torrent’ extension to its URL, and the Miro Team’s Broadcast Machine did all the torrent generation server side, on your own server.

However, Broadcast Machine is no longer supported, and S3 leaves you needing to create an RSS feed. I’ve therefore been busy implementing a bittorrent server that can be hosted at any domain that supports PHP and MySQL. If you can install Wordpress, I intend this software to work too. It can create a .torrent for any file already on your server, and then include it in an RSS based channel feed, ideal for clients like Miro.

I’m looking for a few beta testers to see if this software works outside of my own test set up. If you want to host some video content (perhaps you already post them to youtube) on your own domain, and you are familiar with installing software like Wordpress there, you would be an ideal candidate. It would also be useful to see how multiple downloads work, so if you already use a torrent aware RSS reader (Miro is one), or would like to, please get in touch.

Feel free to email me (john@mcaleely.com), or post a response to this blog entry.

Incidentally, selling this software is not my business idea. This is BSD licensed open source.

I recently rummaged through my loft, and found my collection of Psion PDAs. I’ve got fond memories of these, and there is no way I would part with them. Or so I thought. On review, I had no less than seven of them, including many duplicates.

So, its off to eBay with some of them:

• A netBook
• A Revo
• A Series 5
• The PC Card adaptor for the Series 5

If you fancy buying one, you could always install some of my software!

You can tell I haven’t cycled a lot this spring. It’s taken quite a while to go from twice to three times. It took 1 hr 50 mins to complete the set yesterday.

As the second in an occasional series, Philippe sent me this pic he took. It’s a rather nice image of cyclists showing at the Catto Gallery recently. It looks like I’ve missed the chance to see it for real. Ah well, thanks Philippe!

Other internet video products have frustrating limitations which mean I don’t use them very much: Perhaps the files can only be watched for a few hours; I’m limited to watching programming that can sustain a large technology infrastructure or I have to put up with muddy, low quality, pictures. Often they are just incredibly complex combinations of technologies and applications.

Miro eliminates all these gripes, and I’ve been impressed by the experience of using Miro on my TV over the last six months or so. I’ve tried to write up some of the reasons from a professional perspective over the last few blog entries: Miro respects my own honesty; respects the authoring effort of the people who create video; allows people to publish at low cost and is easy to understand and use.

There’s another reason I like it though - it has become easy for me to download and watch some really high quality video on my TV I wouldn’t see elsewhere. Right now, Miro claims to have more HD content than similar competitors, and it looks great!

I’ve not seen a better way to watch video on the internet.

Miro makes Channels a central part of its user experience, using the same technology as a blog. Having chosen not to offer instant playback video (unlike, say, YouTube) I think that the metaphor of a channel to helps me to understand what I will see after waiting for a download to occur. The TV channel metaphor was easy to grasp, since I’m often asked to subscribe in some way to a channel on my TV. Using the technology behind a blog (syndication feeds like RSS), it makes me consider a flow of new videos in the future, again like a TV channel. By using bandwidth friendly downloading, it can allow a channel to make the promise that the download will be better quality than a live feed.

While it is running, Miro is capable of downloading in the background without impacting the playing video. Therefore it’s often my experience that once I’ve watched one video in a channel, the next has downloaded. By leaving Miro running on the computer attached to my TV, there are also always downlaods that have completed while I’m away. As such, whenever I want to watch channels in Miro, there is something to view. In practice, this means that when I encounter a new channel, I hit the subscribe button, anticipating I will be able to watch it next time I sit in front of the TV.

As an aside, downloading before viewing has another advantage - playback is far more reliable. Most of the internet live feeds I’ve used experience breakup and dropout, making them unpleasant to watch. Miro never suffers from this, which is great.

Whilst I’m generally sceptical about copying the user experience from traditional TV’s and HiFi’s, the channel as blog metaphor seems to work well for Miro.

Hosting video (or any large file) on a webserver has a problem - when a moderately large number of clients want to download it simultaneously the overhead of each individual connection can overwhelm the server. The number of people that can be serviced is quite likely to be lower than the audience for any popular file, and waiting an unpredictable time for the server to become free is a frustrating experience for the viewer.

There are several possible solutions to this:

• Multicast IP
• Hosting on multiple servers via Mirroring or a Content Delivery Network (CDN)
• Collaborative (peer-to-peer) downloading

Multicast IP allows the server to share a connection among many clients, but for various reasons is not available on the public internet so needs to be discounted. Hosting on multiple servers is straightforward, but is costly and out of reach of individual authors.

Collaborative downloading allows the author to require each viewer to also act as a server for that file. In this way an ad-hoc multiple server system can be created, with costs split between the author and each viewer. Whether your viewers will be happy with this probably depends on what you charge them, and what they get in return. Most pay-per-month internet connections allow them to make a useful server contribution at no direct financial cost to them.

A popular collaborative download protocol is BitTorrent, which is included in Miro. An author distributing video to Miro (or similar) viewers can therefore choose whether to deploy multiple servers, or rely on viewers to collaborate as downloads occur. The ability to broadcast without investing in a network of servers should make it materially easier to distribute video from small providers on the internet, in a similar way that blogging has brought publishing words within an individuals reach.

Most of the widely used media players were historically produced to showcase the videos produced by the creator’s authoring tools. They usually recognise that other formats exist, and provide some means of plugging these in, but very few attempt to include all popular formats at once. This has the effect that sometimes, when pressing play, I’m presented with a dialog that announces the file is in an ‘unsupported’ format. While they often offer to download a suitable player, I want to watch a video, and I’m not disposed to buying and installing software. I may not even have the ability to do so. As such, most players don’t work well with all formats and I have to choose a player to match a format, although I’m not particularly interested in the format a video is in - I just want to watch it.

Artists, on the other hand, care a lot about the formats used, as they create the visual limits the artist must live within as they create video. All practical formats create visual artefacts when they are pushed beyond their intended limits, which artists must carefully avoid. As an example, some are good at cartoon content, while others are better at fast moving sport. Another important point is that encoding in a particular format is an expensive operation - video format owners typically charge a lot more for the encoder than the decoder.

Miro attempts to include as many codecs as possible, and has a catholic taste in video formats. Rather than focussing on any one codec, or graciously permitting me to add more, Miro uses a video playback subsystem - VLC - that attempts to play anything you throw at it. I’m not annoyed by strange dialogs that announce they first need to download a codec, and artists are free to select among a wide range of codecs for the one that best suits the video they create.

Including many codecs is not the only option - services such as YouTube instead provide one codec, and re-encode all incoming video into it. Whilst pragmatic, this is sure to lose quality. Any digitisation of a performance necessarily involves making decisions about what to keep, and what to throw away. Of course, the highest quality options produce staggeringly large files that are essentially impractical to use on today’s computers and networks. Whilst a lot of the decisions about what not to keep can be made by the encoder, not all of them can be automated, and high quality encoding is more of an art than a science. As I’ve noted the encoder creates the visual limits for a file, and all of the major ones have various options that can be tuned by a careful operator.

Any translation between modern video formats will involve a loss of quality, and visual artefacts appearing. Whilst I’m often grateful this means I can see any video at all, I’m not happy about throwing away all the effort of the original artist. There are interesting risks - who’d want to watch a football clip where the ball has disappeared due to the automated translation?

By allowing Miro Channels to contain any format of video, the artists can choose the codecs that best suit their purpose. They can also publish the content they’ve already created, without having to re-encode it, opening low cost access to their whole portfolio.