Computing and physics data challenges are distinguished only in Alice.

DC 10/2003 - 09/2004
  Classic one: simulation, reconstruction

DC 10/2004 - 09/2006

Initially we supposed
  
            10% reconstructed data
Tier0 data ----> Tier1's

It has disadvantage that supposes a huge Tier0 (financial problems, etc.)
Now we propose to ship the raw data and make the reconstruction in
Tier1's. This will require very high network bandwidth. 

The size of Alice Computing should not exceed those of ATLAS and CMS.
The size for DC should lead to the final system. The planning exercise was
done by Alberto.

2006 - bye equipment for 2007  20%

2007 mostly proton beam, less data => no need for the full system

2007        -//-         2008  30%

Estimate the multiplicity of the heavy ion events and the need

2008        -//-         2009 100%

Yves presented the document

The physics objectivities were not completely met:
  - simulation->reconstruction chain (not for all detectors)
  - simple analysis tools - not available for most of the tasks
  - digits in Root format (done)

The number from the last production:

 - the size of the simulate events is much bigger than what we
   reported in the Hoffman review. The reason is that we keep hits,
   clusters, etc., which will not be there for the real
   experiment. During the next data challenge we have to decide what
   do we keep. The simplest possibility it to use temporary hits and
   then remove it. 

To be done (for the next DC):
- introduce VirtualMC
- move to NewIO
- introduce the calibration
- implement the ESD classes
- add trigger classes
- make internal planning

POOL - RTag on the persistency (IO)
 - the basic idea was to use Root, but hiding it via abstract
 interfaces. The first version should be released... 

BLUEPRINT - RTag on the global architecture of the LCG (LHC Computing
Grid) software. Proposes the following: 
 - Python as a scripting language
 - ANAPHE as an abstract interface to Root (and to Lite histograming
 package, using ZEBRA) 
 - Root for IO

There was a discussion on the Alice position.

Simulation RTag: VirtualMC and the geom. modeler were presented. ATLAS
is interested in the virtual MC.