In an ideal world (as far as I can imagine it), we would be able to throw a PDO at a URL and have it import the data to i2b2. If we have to have some kind of middleware layer or a bespoke i2b2 cell for this is an open question. In the first instance we just need to be able to talk to i2b2 with a web service.

Both the workbench and the web client talk to i2b2 with web services. Each message consists of a standard xml file with well defined headers containing all the metadata, then any PDOs are included in the body of the xml file. Different parts seem to be able to talk directly with different cells - so the ontology tree display talks with the ontology cell etc.

The web client is written completely in Javascript so it is entirely client-side code. In theory picking it apart would give us a rough idea of how messages are passed to/from i2b2. It is difficult to understand however...

The workbench works in a similar way, but written in Java.

Clearly other groups have got on top of this as they have lots of data in i2b2.

There are a few projects around that look like they might be helpful in pointing us in the right direction.

• ​http://code.google.com/p/o2i2b2/
• ​http://orbit.nlm.nih.gov/resource/r-engine-cell-integrating-r-i2b2-software-infrastructure