Dion Moult Seriously who ever reads this description.

Plans for E2-Productions.com to turn into a personal cloud?

Alert! Alert! Buzzword! Yes, before we start, let’s clear up with what I mean when I say "personal cloud". A personal cloud is a web-accessable system which centralises the function of common web 2.0 services, which may or may not be social. For those that aren’t familiar with this jargon, web 2.0 services are those such as web email clients like GMail, photo sharing and management sites like Flickr, online radios like Last.FM, and even blogs just like this one. So your personal cloud is a system on your very own website, with a web interface for your very own emails, PIM (calendar, notes, todos), images, music, etc. Note that the social attribute is optional. Clouds do not have to necessarily have automatic synchronisation, nor does it have to have the ability to easily share your data with the public.

A little history first. E2-Productions.com used to be the center of attention – thriving with the latest adolescent community fads such as animation, art and music "portals". It later saw the rise of the Blender Model Repository, a personal portfolio, several forums (trendy, weren’t they?), and finally ended with the death of the original animation portal. thinkMoult then emerged somewhere on the Blogspot blogosphere and went on to become a moderately-hacked install of WordPress on the E2 server. E2-Productions had become, and still is now, a dead site.

I’ve been toying with the idea of turning E2-Productions into a personal cloud for quite some time. It did actually occur at one point. Even though a PHP developer myself, it would take too long to create my very own cloud to implement existing free scripts. In the end I had created a network of individual PHP scripts giving me a web based RSS reader, filebin, imagebin, and a proxy. With the helpful addition of several rsync scripts and sshfs, it was usable and offered all the functionality. However of course it was very hacked together and didn’t offer the sort of integration I wanted.

Today I decided to look into other cloud services. One currently in development is ownCloud by Frank Karlitschek, the guy resonsible behind the openDesktop websites. Unfortunately the results were disappointing. Taking into account that it’s still under construction (and therefore incomplete and buggy), like the openDesktop websites themselves, ownCloud is unfortunately yet another developer service that underestimates usability. Whilst it had some nifty features in the works, it’s priorities were skewed away from the rightful mentality that design maketh a website, not the function. Further probing into the code revealed some serious problems with the structure of the coding that didn’t look very well thought out. Needless to say ownCloud is not for me.

The other famous personal cloud is the open-source EyeOS. This one goes all the way, completely replicating the desktop interface in the browser. Again, most of the design of EyeOS 1.x, their stable version, approaches the design of the system as a desktop interface. The canvas of a webpage is not suited for a desktop interface. They both have strengths and weaknesses and unfortunately most of these uncanny designs don’t play to the webbrowser’s strengths. Despite the overblown interface (which excusably is amended quite a bit in their 2.0 unstable version) it’s quite featureful and its extensibility in terms of developing it yourself is quite attractive. However the lag which accompanies such client-side interactive bloat (and server-side too!) doesn’t exactly make it the most practical of choices. It’s definitely worth keeping an eye on, though (excuse the pun).

Further searching yielded an exemplary system called Tonido. Unfortunately due to its proprietary nature it’s not for me either. However it does provide a fine example of the potential of a well executed cloud service. This motivated me to reconsider creating a cloud. I began with considering the basic user objectives for the web interface:

You see the "cloud" I had described (in terms of its most basic user-side functionality) as such is pretty much just a smart web-based filebrowser with mini-webapp additions for (mainly) PIM-data. Our latest newcomer to the fad, Ubuntu One, actually provides these needs very well – leaving the browser to do what it does best, and the rest to the desktop. However it falls short in a vital area – proprietariness. This isn’t a question of evangelism, it’s more of one of the simple requirements of a personal cloud. If we analyse the 5 basic user-side needs to its roots, we get a shorter list of what actually makes a personal cloud, personal:

The first issue is tackled quite well in existing cloud providers, with probably the best implementation (in my opinion) being Ubuntu One. Convenience is another easily satisfied need, with wonderful tools like rsync, sshfs, and version control software (though again, most providers lock you into their own system, and convenience ends once you leave it). However the key feature that in my eyes hasn’t yet been solved by any provider is timelessness. Any proprietary client or syncing software is instantly disqualified due to dependency on the vendor. Now, even if the software was completely open and extensible, many so called personal clouds are simply connecting to existing external services. Whilst consistent with the definition of the personal cloud, the service centralises in what is hoped to be seen as the path of least resistance – that is, only for people actually already using those services. What services am I talking about? Oh, things like Flickr, Google, Facebook and Twitter. Social is a hot topic, but not for everybody. Services don’t even have to be online – dependency on, for example Tomboy Notes or Evolution in Ubuntu One qualifies as a dependency and thus means the service is not timeless.

So how is this overcome? Not easily, for sure. I want to propose that the ideal personal cloud be one that focuses simply on file management and synchronisation. It should stop there – the actual display of files and searching of files should be handled by plugins. Plugins can decide how to properly format an image gallery. Plugins can decide how to display PIM-data. These plugins should be accomodating for the most timeless format ever, plaintext, as well as industry standard formats. The user then only applies the plugins that fits their exact workflow, if necessary writing their own for interpreting their own files. This satisfies the 3 criteria of a personal cloud.

I’m still coming up with a few plans and extra ideas on how it’ll be – but meanwhile I’ve got exams to get over. If anybody knows something similar that exists, let me know, otherwise this’ll be a fun project after WIPUP reaches stable.


hangfire says: (14 May 2010)

There’s been a guy in #gentoo-sunrise working on an ebuild for the Ubuntu One client for the last few days – why not give him a hand?

Dion Moult says: (14 May 2010)

Well, from what I understand, the Ubuntu One “client” is well-integrated in GNOME. As a KDE user, I don’t know what the Kubuntu folks use but I assume it’s going to come a tad later than the GNOME-based ebuild. Anyways, I’ll watch and when it comes out I’ll try it out.

But nevertheless, Ubuntu One still seems to be very dependent. The first issue is that it isn’t on your _own_ server, but instead on some rented spot with Canonical. Also, the features well-describe immaculate integration with things like Tomboy Notes, Evolution and Rhythmbox, something which doesn’t match the “timelessness” criteria I mentioned (and all programs which I don’t use, anyway).

Leave a Comment