Close this search box.

Setting up a framework for a Python implementation of the Open Grid Protocol (technical)

As you might know, the Open Grid Protocol which is worked on by the Architecture Working Group (which consists of people from Linden Lab and the Second Life community) is for virtual worlds what DataPortability is for Social Networks these days. Just done properly from the start ;-)

And the protocol is making progress. Login is defined, many discussions are held about how to send events from the server to the client via HTTP (more about this in another post) and the process of teleporting is also taking up shape. Moreover there was talk about creating a test harness to test individual OGP implementations against a fixed set of tests to ensure interoperability. And in my head there have been some ideas on creating a new sort of libsecondlife in Python but for the Open Grid Protocol.

So I felt it is time to work a bit on the python side of things and did the following:

  • I created an experimental structure for the OGP library and an agent domain.
  • I tried to get the new OGP login to run as it already is experimentally implemented by Linden Lab on a test grid.
  • I tried to get the capsserver from Linden Lab to run which is included in mulib (more about Capabilities here).

Success is mixed. I wasn’t able to get login to work (but now get an error message telling me that authentication failed instead of just a 500 error) but therefore created the structure for a library and an agent domain and got the Caps Server to run.

I put all this into an „pyogp“ namespace and put this into my pysecondlife project on Google Code. The pyogp package might be moved into it’s own project though and I also need to discuss with Linden Lab how they want to handle the whole thing. I mainly wanted to get started with something. You can find more information on the pysecondlife Google code page. It also explains how to get the caps server to run. The only thing you need should be to have Python 2.4+ installed.

The next steps should be to discuss this with the Architecture Working Group and to get some people involved. From a technical viewpoint I would like to:

  • implement login for the agent domain
  • implement a login for the library to connect to it
  • create a test harness and experiment if it’s better than with or without the unittest/doctest framework

If you are interested feel free to join us at the next Architecture Working Group(ies) meetup on tuesday. All the details can be found here (here is the Google Calendar for it). Next tuesday we will also have Donovan Preston talking about RHTTP to make event queues working via HTTP.

Teile diesen Beitrag