Regarding the EF-infrastructure activities, you can find below a brief 
wrap-up of the major development issues/plans. The EF infrastructure 
activities are divided in 2 categories: DataFlow infrastructure and HLT 
infrastructure. I asked Serge to fill the list for the second part. 


A) EF: DAQ infrastructure 

Adopt tdaq wide "Error Reporting System" in EFD;
migrate from the custom EFD implementation to the ERS one or put ERS on the
top of the former [timescale: tdaq-01-07-00, FTE: less then 1 month] 

Improve EFD backpressure mechanism;
absorb rate oscillations inside the EFD preventing the propagation to the
Event Builder system [timescale: asap, FTE: 1 month, but Sander is already 
doing that] 

Make the event format library aware of the EFD sharedHeap.
This is a complex issue. I try to summarize it in a few lines. During their 
transit inside an EF node the events are stored in a shared memory mapped 
file (called sharedHeap) managed by the EFD. A PT receives read-only pointers
to the raw events and therefore it cannot corrupt the event. In the current
implementation, the PT can "add" a subDetector fragment (containing
reconstruction information and serialized objects produced by the PESA 
selection algorithms) to the raw event. I quoted the word "add", because 
the PT does not touch the raw event, it simply creates an eformat subdetector 
fragment, ask the EFD for a writable sharedHeap zone and serialize there the 
subdetector fragment; the EFD is taking care of appending the subdetector 
fragment to the raw event. Actually also the EFD do not touch the raw event, 
it exploits the EFD virtual event facility: the raw event is untouched and 
the changes to an event are stored elsewhere (a la "cvs"). When the event 
is sent to the SFO, the "changed" event is serialized on the fly on the 
network.
Now, we would like to exploit the virtual event concept, allowing the PT to 
do (to "steer") complex on the event (not only add a subdetector fragment). 
E.g.: a calibration PT should be able to strip the event (remove useless 
fragment). In order to do that the eformat library (used by the PT to 
access/manage the event) must be aware of the EFD sharedHeap and the EFD
virtual event concepts.
[timescale: tdaq-01-07-00, FTE: 3 months (??)] 

Optimization of the EFIO protocol 
The EFIO protocol has been changed in order to fully exploit the gigabit 
network and some tests on preseries are needed [this is a "minor" activity 
almost covered] 

B) EF: HLT infrastructure

development of EF internal routing and generalization / improvements for 
Streaming, with Calib. algo: 
a) generalization of EFResult class (to contain both/either PESA or Calib 
payloads in EFResult objects) -- (see attached slides) -- [already ongoing, 
people in place, expected done Sept. 2006] 
b) generalize routing inside EF (part. DF) according to configurable 
mapping mechanism (see pages 10-13 of attached slides), work together with 
DC [ongoing, people already allocated, expect done autumn-winter 2006] 
c1)  create Calibr test algo 
c2) and test generalized EF-ROUTING with this test Calib Algo 
c3)  and perhaps make yet another test Calib. Algo with DIRECT writing/updating
Calibr. DB and make DB-algo-performance tests with this algo. 
[All these (c1, c2, c3) are important tasks and connected both with (b) and 
with activities on detector calibration. No particular people allocated right 
now and is good topic to contribute to. Expected time connected with (b) and 
LST ] 

EF infrastructure updates for monitoring functionality 
already mostly done,  just support needed in infrastr. (people available) 
and particular USE applications on side of Detectors (monitoring of HLT 
operations as client here). Good contribution could be developing "client-type"
(plugin-type) HLT-specific application for monitoring of HLT operations in 
Atlas Control Room - discussed, need to go in contacts with Monitoring Working 
Group, but no people allocated yet. Timescale - start now, ready by LHC the 
sooner the better for testing. 

testing: HLT commissioning in P1
not enough people available yet, only goes partially and by whoever available 
in short periods, would need permanent assigned/responsible person, but based 
in cern (because it requires personal presence in P1 and to many 
meetings/discussions) 

testing: Large Scale Tests
SLAC group already involved in DB-replication for HLT and, of course, 
additional contribution in this area would be important 


So the best potential area to contribute would be A3, B1c and B4 and 
obviously general discussion about streaming. 

Sorry again for answering only now!
Let me know if clarifications or further details are needed.

Regards
    Andrea