An excellent in the wild post written by Josh Patterson about his floe.tv project with some feedback from me, so the credit really goes to him. For those of you unfamiliar with this series of posts, the idea is to create blog-length interviews with various in the wild apps describing their processes and the technologies that they use with regards to data portability. The goal is to profile real use cases, solutions, and lessons learned when it comes to the current state of affairs for data portability technology. Note that these posts aren’t meant to recommend or not recommend certain technology, I leave that up to the developers/architects to decide based on their needs. If you have such an app and are interested in being interviewed, please leave me a comment on one of my posts and I will get in touch with you.
Basically with floe.tv we started out just wanting to try out some video ideas we had. We started out with a simple playlist of videos on the internet linked together at a site, which evolved into a full own video editor. After a number of months of development we hit a point where the team sat down with some users and did a testing session, asking questions and gauging responses to see how well we were hitting our marks. We came to the subject of data storage, local hardrives, and getting media online, and just as a thought exercise I asked “well, what if floe.tv just knew about all your online media by your login name, and referenced it automatically in your libraries the first time you logged in — just as if it was an app installed locally on your hd?” and immediately both of them became excited and one asked “can I do that right now? when can I use that?” and I knew from experience that the market was speaking very loudly and clearly in my direction, and that I had better listen very closely.
The very next meeting I posed this question to our team:
What if our app was “inherently installed” in the internet? What if someone logged in, and the app just acted like a desktop app that “knew” about your flickr images, your youtube videos, it knew about your myspace friends, facebook friends, and automatically treated them as one logical database, one logical social graph? And someone started right into an app tutorial right off the bat with their contacts, files, and assets already referenced (but fully respecting privacy, control, etc)?
So the next question naturally becomes “that all sounds really great, but … how do we get there?”
From there we really began to push a “what if this/that” scenario, and drew up our ideas into a document entitled WRFS and from that we began to re-engineer the floe.tv app towards a truly linked data experience.
We are currently using FLEX/as3 for the editor and player with ASP.NET for the server technology. Discovery is a big deal with how I view next gen web apps — dynamically finding data at runtime without having to go through “Big Data’s” walls to get to it. How I see this happening is
the user logs in with an identity url, be it an openID or a XRDS file location. floe.tv is an openID only application, but user’s can map multiple openIDs to one account with it.
the app authenticates the user, and then uses multiple fallback methods to find the XRDS-S file if not specified.
The XRDS file is parsed, and relevant data types (here image and video) location uris are pulled out
- Each uri points to a data container, which is then queried via its API for a list of resources the user has stored there.
- these results are then aggregated back together into a single “recordset” to be returned to the floe.tv application layer
The fallback methods for XRDS discovery (done in the FLEX client) are ordered as:
- First do basic yadis discovery to see if the identity delegate is the openID provider or a blog of some sort head link delegation setup. Either the openID provider or the XRDS location may here, or both. In some cases, such as the DiSo XRDS plugin, the XRDS file located in the head link tag will have the openID provider location.
- The secondary method we have kicked around is to query the openID provider for an Attribute Exchange key that points to the XRDS file. This is not well defined yet but has been discussed amongst various groups.
- Lastly we fallback to having our flex app prompt the user for a XRDS url so that we can “enhance their user experience”.
So although I think our secondary option with the AX key is a little shaky right now, overall we degrade gracefully.
For a quick demo of how some of the data query and aggregation mechanics might work, I’ve built a short demo illustrating the step by step mechanics.
We aren’t done with this application, obviously, and a lot of work remains to be done – I should note that I am currently at Step 2 of 5 as stated above. However, the application is evolving into the embryo of what I think a linked data application can and will be. What I can share are the places that we are actively looking for solutions, simple decentralized solutions, that solve these issues.
One thing that is slowly changing is the perception of cross domain scripting on the internet. As services get more complex, and require more aggregation of data from multiple sources, we are going to push more data handling duties to the client, as scalability will suffer. For flash crossdomain scripting we need the crossdomain.xml file present on the server of the api would like to call for it to work. This is a trivial thing to setup as it consists of a simple xml file located off of the subdirectory level you wish to grant access to.
Once this file is exposed, the flash runtime will then allow calls from the client to those servers.
We’ve been waiting to see how the discovery wars pan out, as it stands now XRDS-S is looking like the service index of choice amongst the big players (presently Yahoo is endorsing it, hiring Eran Hammer Lahav as their Open Standards Evangelist. How the XRDS resource is discovered automatically without a tremendous amount of user interaction is something that we are taking many approaches towards, as discussed above. For now we’re going to focus on finding XRDS files as our catalog of service endpoints for a user. The DiSo project is going to be publishing its user’s service endpoints in the XRDS format and already has a plugin for it, so I think in the short term we’ll be focusing on consuming that data in terms of an early conceptual demo in runtime linked data.
Once we find the XRDS file, we aren’t out of the woods. How do I set my XRDS file up so that I can tell the floe.tv application that I have “images in flickr”? This is a fundamental question being setup and worked on at the site http://www.xrdstype.net/ by many different groups and people, and has yet to fully be fleshed out.
There’s also the issue of once we find a data endpoint, how do we talk to it? ATOM? Some sort of standard data interop api that spits out RDF? In a perfect world I’d love to see a self organizing web, a linked data web that can find a data endpoint at runtime, find its semantic schema, and wire itself such that it can talk to that api without ever needing user intervention — it simply will understand how to talk to it.
Another key development will be the permission system, possibly using OAuth+Discovery to automate the updating of a XRDS-S file for someone when they add data to a service. I need to learn a lot more about OAuth and the direction its heading. I’d prefer a world where the user doesn’t need to manually go to a site to allow their resources to be used by a third party, but for now, this is how we have to operate.
So really, I think I have more “lessons underway” than “lessons learned”, but sharing this information is key since it sparks interest in other like minded developers who may know a lot more about some of these areas than me. There’s going to be some places that we punt and just hard code some scaffolding code in to just get going, but over time I’d like to evolve towards a linked data web that auto discovers new connections at runtime and self organizes to give a smarter and far more intuitive user experience than we’ve seen so far.
If you are interesting in linking data or trying some data interop experiments, please feel free to email me (jpatterson at floe.tv) or check out the WRFS workgroup or my blog http://jpatterson.floe.tv
Feedback and suggestions are welcomed.