[Dev] PyCon 2005 report

[Ted Leung]

Here's my take on PyCon this year...

== Sprints

This year's sprints were an unqualified success.

=== Successes

We had three non-OSAF folks work on two parcels (one only stayed for  
one day), and got them working.  There were two groups sprinting on  
parcels.  Alec Flett, Jeffrey Harris, and Bill ??? worked on a Flickr  
parcel.   Mark Hertel, Kragen Sitaker (1 day only), Katie Parlante, and  
I worked on the del.icio.us parcel.   The parcels worked well enough  
that we were able to show them during our presentation in the  
conference proper.   This was a big bonus -- we'd been trying to write  
the paper to try to convince people that we had enough built so that  
people could build parcels.  When the sprint parcels got done, it was  
easier to show proof than to tell.

Brian Kirsch spent a ton of time with the Twisted folks (as noted in  
his post to dev@).  The Twisted folks have given Brian and Bear commit  
privileges to Twisted, which I consider a milestone in our relationship  
with them.

=== Difficulties

We discovered that there are a number of sharp edges that will cut  
people trying to develop parcels.  Many of these are related to the use  
of XML to describe Kinds, and to the fact that people have to update  
the same data in multiple places.

We had to disable the WakeupCallers that got written for the sprint  
parcels because of some bugs.

Sprint page: <http://wiki.osafoundation.org/bin/view/Journal/PyCon2005>
Feedback page:  
<http://wiki.osafoundation.org/bin/view/Journal/ 
PyCon2005SprintFeedback>

One day during the sprints Jeffrey and I were trying to explain what  
Chandler was to two folks who had never heard of Chandler.  After they  
took a look at the website/wiki, they said "I still have no idea what  
you guys are doing".  Fixing this is important.

== PyCon proper

=== Presentations

* Chandler - Our presentation was moderately well attended.  We were in  
the largest room, which I recall being somewhat less than half full.   
We had a few technical glitches at the beginning, but after that things  
went pretty smoothly.   We ran out of time to take questions, which is  
how you want a presentation to end.  The Chandler Open Space (BOF)  
lasted for about an hour later that day, and had sizable attendance -  
around 15-20 people.   People had lots of questions, and it seemed like  
there were some people who were interested in downloading 0.5 and  
trying to build some parcels.

* PyLucene - I missed Andi's presentation, but I heard PyLucene  
mentioned a few times during the conference.  It seems that

=== Conversations

Here are some relevant blog posts:
     <http://www.postneo.com/2005/03/24/pycon-dc-2005-day-2>
     <http://yergler.net/blog/archives/2005/03/25/pycon-wrap-up>

Brian has already related Divmod's interest in using Chandler instead  
of building their own rich client.

The organizers of EuroPython tracked me down and asked me to come to  
EuroPython and give a presentation on Chandler.

Abe Fetting, the author of Yarn is enthusiastic about us using Yarn in  
Chandler.

== Action Items/Implications

I think that it was definitely good to have a larger presence this  
year.  It is clear that PyCon is the conference to come to in order to  
get work done and integrate better into the Python community, and a  
larger presence means more opportunities.   The talks are also much  
more relevant to developers working in Python.

PyCon was the first time we were able to show anything that was  
remotely like what we intend to build.  Chandler 0.5 shows enough of  
the big ideas (ItemCollections that hold multiple kinds, Items in  
multiple item collections, stamping, and parcels), that some people  
were able to get an idea of what we are trying to build.  Not everyone  
saw that, but the people who did were excited by that.  This was an  
important first.   I think it was also important to be able to  
demonstrate running code (both our own, and the sprint parcels).

=== Projects that warrant further investigation:

     The MDA projects mentioned by Phillip, in particular Traits and  
Envisage (I exchanged cards with the Envisage speaker)
     PyICU
     Yarn
     py.test
     PythonEggs - http://peak.telecommunity.com/DevCenter/PythonEggs -  
This should be easy.

==== Developer platform improvements:

Based on Sprint feedback, here are some short term and long term things  
that we need to do in order to improve the situation for parcel  
developers:

Short term

* Document the various repository types that are available
* Document the kinds that exist in the system and give guidelines on  
how to derive from them.
* Document the various kinds of blocks that exist so that someone can  
make a detail view without hand holding.
* Document the use of the web-server based repository viewer as a  
debugging tool
* Fix any bugs uncovered by the sprint parcels
* Decide whether we want to include the sprint parcels and how

==== Longer term

* Remove parcel.xml and replace with a system that uses Python classes  
as a much as possible (ideas from Spike should help here)
* Flatten the module hierarchy for parcels (at least)
* Simplify the XML format used to describe blocks/the detail view.
* Improve the error messages and diagnostics issued by the system
* We need support for preference panel type functionality
* Figure out how to change/update the website/wiki

=== Other Actions

* Decide whether to accept the invitation to EuroPython

----
Ted Leung                 Open Source Applications Foundation (OSAF)
PGP Fingerprint: 1003 7870 251F FA71 A59A  CEE3 BEBA 2B87 F5FC 4B42