I do my own stunts

Fabricio Zuardi's weblog, English version.
Tue Nov 18

Internet Platforms revisited

On September of 2007, Marc Andreessen wrote an interesting blogpost about internet platforms with some good observations and a preview of what could be the next generation of internet platforms. The 2 most important points from that post that I would like to repeat here are:

1- If you can program it, then it’s a platform. If you can’t, then it’s not.

And:

2- the difference between the chicken and the pig at a ham and egg breakfast. The chicken’s involved but the pig is committed.

At that time, Ning itself was one of the companies working on a model that was close to this ideal “Level 3” platform… but that’s not the point of this post. Suffice to say that due to market-fit and growth we are now back to the transition to Levels 1 and 2 , which is not a bad thing, there are plenty of successful business built on top of platforms of all levels, as Marc said back then, “they’re all good… …Having a platform is always better than not having a platform, period.”

This post is to showcase 4 examples selected by someone who codes (myself) of companies that are currently getting close to that ideal of the “runtime environment”, the “committed platforms” with “develop in the browser” tools, infrastructure to “run in the cloud” and if possible with an “open source ecosystem within the platform to let users freely share code with one another”.

The platforms are: Freebase Acre (aka freebase applications); Google AppEngine; Heroku; and Appjet.

Freebase Acre

  • Applications are programmed with server-side Javascript, a template language that extends html and MQL
  • The interface to edit the source code is web based
  • Users can view the source for any application, and clone them easily
  • The application is hosted by Metaweb
  • The application runs in the cloud
  • The application has to run on Metaweb’s infrastructure. If the pig becomes a chicken you are out of luck

Google App Engine

  • Applications are programmed with Python(a limited version of it) and an API to the “DataStore”
  • There is no web-based interface to edit the source code, you have to develop/test it on a local SDK and then deploy
  • Users cannot view the source for any application, all apps are closed by default
  • The application is hosted by Google, however, the SDK(a limited version of the platform code) is open source.
  • The application runs in the cloud
  • If the pig becomes a chicken, migration is kinda-of possible (because a version of the platform code is open)

Appjet

  • Applications are programmed with server-side Javascript(with small extensions and a mini-framework with helper functions)
  • The interface to edit the source code is web based
  • Users can view the source for any application, and clone them easily
  • Modularization and KISS is encouraged (1-file apps + make-it-a-library mentality)
  • The application is hosted by Appjet, however, the necessary files to run an appjet app runner in your own server are available and open source.
  • The application runs in the cloud(if you want)
  • If the pig becomes a chicken, you can migrate your app to another host (because a version of the platform code is open)

Heroku

  • Applications are programmed with Ruby/Rails
  • There are web based tools to edit the source code, but also git integration if you prefer to develop locally first
  • Users can view the source of apps flagged as “public”, and clone those is easy
  • The application is hosted by Heroku
  • The application runs in the cloud (using EC2)
  • Since it is a Rails app, you can easily migrate your app to other hosts if the pig becomes a chicken
Comments (View)