Index: branches/eam_branches/ipp-20101205/psphot/src/psphotStackReadout.c
===================================================================
--- branches/eam_branches/ipp-20101205/psphot/src/psphotStackReadout.c	(revision 30023)
+++ branches/eam_branches/ipp-20101205/psphot/src/psphotStackReadout.c	(revision 30027)
@@ -2,4 +2,5 @@
 
 # define STACK_RAW "PSPHOT.STACK.INPUT.RAW"
+# define STACK_CNV "PSPHOT.STACK.INPUT.CNV"
 # define STACK_OUT "PSPHOT.STACK.OUTPUT.IMAGE"
 
@@ -20,8 +21,10 @@
     PS_ASSERT_PTR_NON_NULL (breakPt, false);
 
+    BST = RAW : CNV;
+
     // we have 3 relevant files: RAW, CNV, OUT 
 
     // set the photcode for each image
-    if (!psphotAddPhotcode (config, view, STACK_OUT)) {
+    if (!psphotAddPhotcode (config, view, STACK_BST)) {
         psError (PSPHOT_ERR_CONFIG, false, "trouble defining the photcode");
         return false;
@@ -190,2 +193,97 @@
 }
 
+/* here is the process:
+
+ * we have three(*) images:
+ * RAW : unconvolved image stack 
+ * CNV : input convolved image
+
+ * OUT : psf-matched output image (there may be more than one of
+ * these.  we will generate the first matched image by selecting the
+ * target PSF and doing a full psf-maching process (as used by ppStack
+ * and ppSub).  But, additional target output files should use a
+ * simple gaussian convolution kernel determind from therms of the
+ * current and the target).
+
+ * the output should be / could be one of the matched images, but not
+ * all.  should we ensure the first gets written out, and ot save the
+ * others (or only optionally).
+
+ * by default, we probably only sve the cmf ffile outputs.
+
+ * load the RAW image (unconvolved stacks)
+ * add photcode to the output headers / readout->analysis
+ * generate mask and variance image (this is probably never needed in
+   practice: we always load an input mask & var.
+ * generate & subtract a model background for ?? (RAW? CNV? OUT? all?)
+ * load a PSF (probably not yet working)
+
+ * generate the CHISQ image from the RAW input images (why save on OUT?)
+
+ * find detections on RAW
+
+ * copy detections to OUT
+
+ * generate source stats (moments) for OUT
+
+ * match sources across inputs (on OUT?)
+
+ * generate source stats for the new constructions
+
+ * rough class (star, galaxy, cosmic, etc)
+
+ * Image quality
+
+ * generate PSF
+
+ * guess models
+
+ * merge sources (new -> old)
+
+ * linear fit to the psf
+
+ * find ApResid
+
+ * assign common positions
+
+ * radial apertures (** this should be on the PSF-matched images
+
+ * extended analysis (elliptical profile & petrosian)
+
+ * extended fits (sersic, etc)
+
+ * psphot magnitudes
+
+
+ ******
+
+ the above is all wrong:  first, we sohould be doing the full
+ morphology analysis (ExtendedAnalysis & ExtendedFits) on the CNV or
+ RAW image (as desired optionally), etc.
+
+ In the discussion below, 'BST' (best) means optionally RAW or CNV
+
+ * detection : RAW & CHISQ (of RAW)
+ * moments : used by psf analysis & classification (BST)
+ * rough class : uses moments, not pixels
+ * image quality : uses moments as well
+ * generate PSF : (BST)
+ * guess models (BST)
+ * linear fit (BST)
+ * find ApResid (BST) -- uses sources not pixels
+ * extended analysis (BST)
+ * extended fits (BST)
+ * detection efficiency (BST)
+ 
+ * somehow need to copy the sources so they point at the pixels on the
+ * OUT image 
+
+ * foreach target PSF
+   * radial aperture
+   * convolve to next target PSF
+
+   * somehow need to organize the output file to have the values from
+   * the different PSFs in separate tables (with header info to
+   * specify the size of that PSF)
+
+   */