The end of reposiphilia?

One sure sign that Web 2.0 is a genuinely new epoch in the tech industry: the haymaker punch it threw at a repository-centric view of application architectures, effectively knocking it out of the ring. For those who aren't familiar with what I'm talking about, I'll give a little bit of history for the young 'uns out there.

Forward into the past
Let's hop in a time machine, go back 10 years, and eavesdrop on conversations in development teams building multi-tier applications. Chances are you'll hear no small number of words about the repository. For example, suppose the project was integrating two middleware applications, such as content management systems and ERP applications. In many development teams, you'd get funny looks if you didn't advocate some merger of the two repositories as the solution to the challenge. Integration at the middle tier sounded, to many ears, like trying to pull a fast one, substituting a hack for "real" integration.

Of course, the reposiphiliacs had an argument worth taking seriously. If you didn't have a single repository, you'd impose more work on the DBA or sysadmin. Any repository-level rules--stored procedures, data retention policies, etc.--would have to be repeated across every repository. If standards didn't compel people to build similar metadata schemes, you'd spend a lot of your time translating between applications, often with very different notions of the same attribute. (Over here, it's an array; over there, it's not.)

Everything you know is wrong
Now, let's jump back to the present day. One of the striking things about the Web 2.0 world is not only the ease of integration, but also the relative unimportance of the repository in the discussion. You can add your Flickr album to your Facebook profile. Mashups keep on growing, both in options (what you can combine) and methods (how you build them, and the level of technical skill needed). You can upload documents to Box.net, and use Zoho to edit them in place. RSS has grown from simple newsfeeds to an inter-application communication medium. When I publish this post, a link will appear automatically in my Twitter feed. And so on.

Nowhere in this discussion do you hear the word "repository," nor do you need to. Certainly, data structures matter, as do the media for communicating metadata from one application to another. But the underlying repository--by which I mean, if it's not clear, the place where the content is stored--doesn't really matter all that much. In fact, as Web 2.0 charges forward, the repository continues to be commoditized. Amazon, Salesforce, Intuit, or someone else will handle the structured or unstructured data storage and management for you.

What many once considered to be a kludge--integration above the level of the repository--is now the main platform for innovation. Now, if that's not a sign that we live in a different world, I don't know what is.

Comments

re: The end of reposiphilia?

I couldn't agree more with your contention about the "repository" itself mattering less and less. The question where your data is stored has already been answered, for the most part - the exciting part will be the solutions around how you can access content and what you can do with it, regardless of the location, platform or format.

re: The end of reposiphilia?

Integrating repositories may have been a mantra a decade ago, but it doesn't mean it was wrong at the time. There have been a number of changes over the past decade, one of which is the loosening of the definition of "internal" IT.The other big change that has happened, and continues to happen is the change in definition of what an application is and how we use those applications.We're much more comfortable today with loosely coupled applications than we were 10 years ago. There is no single reason for this, but better transport and data security, better bandwidth, more comfort with distributed entities such as SalesForce.com and other hosted apps all make this possible.When tight integration is needed, then some form of repository integration -- even if it via some form of federation -- will have benefits and be needed.Keep in mind that most of the examples you cite, and most of the loose coupling out there is related to content/data sharing. This is easy. A lot of repository integration had to do with metadata (or even meta-metadata) i.e. rules and logic on how to manipulate data (or rules).I'm not sure we're yet at the point that that kind of integration can be done easily via Web 2.0 technologies or mashups.Saeed