I have been investigating the dtime_% fields in the diff and chip stages for gpc1 for the month of October (and also September, see [wiki:ProcessingTimesSept2010]).  The tag changed on 2010-10-29, which included a number of changes to the diff stage.

== Diff Stage ==

There are 4 dtime columns for diff: dtime_match, dtime_script, dtime_phot and dtime_diff.  dtime_diff is never set (it is always NULL).  dtime_match and dtime_phot often have nulls (about 7% of the time). I excluded from my plots diff_ids with NULL values in any of the dtime columns (except for dtime_diff).  For the 'old tag', I selected dates between 2010-09-28 and 2010-10-25, and for the 'new tag', I selected dates between 2010-10-29 and 2010-11-03.  

 * new tag = dotted lines
 * old tag = solid lines

 * Purple = dtime_script
 * Green = dtime_phot
 * Black = dtime_match

[[Image(diff_dtime-10.png)]]

And here are the median times for each:

||dtime || old tag(s) || new tag(s)|| (old tag - new tag)(s) ||
||dtime_match || 54.4 || 40.4 || 14.0 ||
||dtime_phot || 70.0 || 59.0 || 11.0 ||
||dtime_script || 151.0 || 125.0 || 26.0 ||