Changes between Version 7 and Version 8 of UniqueIds


Ignore:
Timestamp:
01/28/2009 06:49:59 PM (10 years ago)
Author:
rhl
Comment:

Comment added.

Legend:

Unmodified
Added
Removed
Modified
  • UniqueIds

    v7 v8  
    102102 
    103103 
     104==== Comment by rhl on Wed 28 Jan 2009 07:49:59 PM CST ==== 
     105I don't see how this scheme handles reprocessing the data with a different 
     106version of the code.  Also, it pushes the responsibility for generating 
     107globally-unique IDs down to the image processing code;  I'd be much happier 
     108generating an ID that's unique within the area being processed, and then 
     109generate the unique ID later.  This is analogous to my intent to generate 
     110fluxes in DN and positions in pixels, not in Janskys and ra/dec --- image 
     111processing code should only generate what it knows. 
     112 
     113If we choose the "first" method things are even worse; I have to use 
     114calibrated quantities (positions) in a pipeline that otherwise only  
     115needs to handle pixels. 
     116 
     117Scheme two would be OK if the image code is passed a unique ID to add to its 
     118per-object IDs --- but even that's tricky when the detection is run over 
     119a number of threads (the "unique ID" couldn't be just the exposure ID, 
     120but would have to provide a separate range of (small) object IDs to  
     121be generated by the detection code. 
     122 
    104123[[AddComment]]