Please add here any topics you would like to see discussed at the DiFX user's meeting.

• The status of DiFX-2.0 and advantages compared to DiFX-1.5.x (Adam Deller)
• Legal characters in sources (WB: i.e., slashes are bad! What else? What maximum length to enforce?)
• Upper/lower sideband cross correlation (WB: combine with discussion of shifted FFT)
• Cluster performance (benchmarking and pitfalls)
• Output file versioning (e.g., managing multiple correlation attempts of a single job)
• Mark5B issues
• difx/mpi setup scripts
• Monitoring (visibilities AND performance)
• (partially-)Overlapped FFTs, both for better high lag performance and to get an integer number of FFTs into an inconvenient subint length (WB)
• Semi-formal “discussion” sessions during the hackathon. Someone stands in fronts and directs the discussion about some new feature etc. E.g.
  • Circular to Linear conversion
  • Broadcast of “useful” monitoring info
• Case sensitivity in vex2difx (CP. WB; also applies to other places within DiFX)
• Walter's list of long term DiFX desires / changes
  • Full VDIF support
  • Support for K5 format
  • Audit of amplitude scaling
  • Within native Mark5: ability for dual-bank playback
  • Documentation: man pages for DiFX programs?
  • Polynomial(l,m) UVW and delay model for wide field case
  • Improved run-time predictor for jobs
  • Investigate whether or not to rereference band frequencies to middle of band
  • (eventual) support for python 3
  • Possible switched-power detection based on autocorrelations
  • shifted FFTs
• Support for clock models and EOPs in Vex file (Cormac)
• Intel Performance Primitives version 7.0: adapting to new installation procedure and directory structure (WB)
• Mark5/cluster configuration (HR,WB)
• JIRA / bug tracking (WB)
• SVN cleanup (CP)
• Generalizing the NRAO system (HR)
• Documentation of mpifxcorr architecture (JM)
  • So that a C++ literate student can dive in and make changes with minimal tuition
• Unit testing (JM)
  • model accountability
  • bug tracking
• Look for overlaps between difx2fits and difx2mark4 and move into difxio (JM&WB)
• difx pcals into FITS (JM)