weak pixels, etc.

[Stupendous Man <richmond@a188-l009.rit.edu>]

  A few comments on the recent discussions of image defects, now that
I've returned to RIT.

    - as Arne says, if you know that some area of the chip has a defect
             which cannot be corrected, it's best to discard any objects
             found in that area

    - or, if you keep such objects, you could mark them as "doubtful" 
             in some way

  The Mark IV reduction pipeline I've written does this: it uses the 
flatfield images to search for areas in which the chip's response seems
far from ordinary, and creates a list of areas to "mask".  Any object
found in one of these areas is marked with a flag in the output star lists.

  I always planned that a database would have a large number of flags
available for each observation, one of which could be set for this reason.
Other flags might be "close to the edge of an image", "close to another
detected star", "has position which corresponds to known variable,"
"shape is unlike an ordinary star", and so forth.

  Maciej has asked questions about errors which occur during his 
attempts to run the pipeline on some data.  I'm afraid that I'm running
into a wall: there are so many possible reasons for the pipeline to fail
that it's very difficult for me to figure out exactly which one is
the right reason.  This will happen when other people run the pipeline,
too.  I suspect it's a classic "user support" problem.  If I were able
to sit down in a room with Maciej, and talk to him about it, and then
watch him reduce some data from the start, it would probably be obvious
what's going wrong.  But when we can communicate only by E-mail, 
it's very hard.

  I don't want to start a discussion on this topic now, but it might
be a good idea for anyone who is thinking of writing any kind of 
software for the project to consider: how will you document your work?
Will you offer support?  If so, how?  How many hours a week are you
willing to devote to support?

  My guess is that somehow, Maciej is running the pipeline in such
a way that some step is being skipped.  The reduced images are supposed
to have sky values above zero.  Some of the star-finding and star-measuring
code assumes a positive sky level.  If semi-processed, or un-processed,
images are fed to the star-finding/star-measuring procedures, errors
can occur; and no stars may be found.

  I just purchased a new 80 GB hard drive for one of my computers
here.  I'd be happy to install a test database on it, once we get 
to the stage of testing remote users' access to the database.  
RIT has T1 lines running into it, and my machines are on 24 hours 
a day, so I can provide a decent environment for people to test their
work.

                                             Michael Richmond