Building Feedly

Some people are surprised to see Feeddo being so far from our infrastructure roots/base. On the one hand, it is true that it is a stretch (and an interesting learning experience). On the other hand, the Feeddo consumer service is only the tip of the iceberg. In this post, I am going to explain how it relates to some of the infrastructure trends we believe in.

Trend #1: The emergence of a new UI/logic abstraction

Every once in a while a new abstraction emerges which divides and refines a problem space: SQL and the relational data model enabled the separation of the data model from the core of the application. Over the last five to six years, they has been a lot of emulation around web services and content syndication. We believe that, as a result, we have reached the point where the combination of REST, JSON and ATOM/RSS is going to enable a significantly better abstraction for separating user interface and business logic. Significantly better because …significantly simpler, …inherently document-oriented and coarse-grained, …supporting both structured, semi-structured, multi-media and micro-formated documents and …language/platform neutral.

Trend #2: Federated data and services

A large number of the existing applications depend on a single data source/database. The reason this pattern has been dominant is that it simplifies transaction processing and management of the application. Integration is usually performed as a back end batch process where data is replicated to other systems.

But a shift, driven by the increasing size of the data, the need for more real-time experiences and the need to integrate with transactions systems which can not be replicated, is about to happen. You see already a lot of “mash-ups” emerging which try to tie together data and services from multiple back ends in a more dynamic and transactional fashion. The iTunes ATT iPhone activation being one of the most interesting one.

We believe that more patterns and infrastructure components are going to emerge to simplify the complexity of building reliable and transactional applications across a distributed data model and that the abstraction we described in trend #1 will be an important catalyst.

Trend #3: Evolution of business logic

As a cleaner separation emerges between the UI and business logic, the server frameworks will evolve to increase the congruence of the interface required by the abstraction and the mental model of the developer.

Data abstractions will evolve into semi-structured multi-media documents captured in JSON and ATOM. Business Logic will the partitioned into pipelines of -multi-language- interceptors - assembled using declarative policies and business rules - processing business documents - either synchronously or asynchronously. The persistence layer will evolve into a data grid model combining traditional databases, cluster caches (like memcached or tangasol) and storage services (like amazon S3).

Trends #4: Evolution of the presentation/interaction logic.

This is one of the areas where there is still a lot of emulation and it is not clear which pattern(s) might dominate.

We are cheering for AJAX plus and are huge fans of the work firefox and Yahoo have done over the last seven years in terms of defining patterns which allow combinations of HTML (domain model), CSS (presentation logic) and Javascript (behavior logic) and respect the page centric, URL addressable nature of the internet/browser.

Feeddo

Feeddo started as an experience to help us explore, validate and adapt some of these trends: … building the Feeddo UI allows us to understand some of the consistent set of issues developers face when building mash-ups, …building the Feeddo back end allows us to understand the consistent set of issues developers face when building RESTful services driven by JSON and ATOM and finally delivering Feeddo as a service to consumers allows us to understand the operational aspects of the problem as well as a directly link to what next generation information portals might look like.

“If you can’t measure it, you can’t manage it”

This post in a more “infrastructure view of Feeddo” which I hope will initiate some interesting technical discussions. In my next post I will try to talk more about the innovation and design dimensions of Feeddo and what we do to measure those more subjective aspects of the project.