Looked (in Yusra's database) at the overlap region between the NCSA and IN2P3 processings. Surprised to find that, although the processing design was explained to be expected to generate independent sets of IDs in the two productions, roughly 3/4 of objects appear to have ended up with the same IDs.
Looked at differences between the photometry from the two processings in the overlap region, finding statistically significant differences in g-band photometry at the ~0.01mag level at the bright end and ~0.05mag at the faint end.
Was unable to use the UI effectively for test cases that involved retrieving photometry (including light curves) for a set of objects from an external list.
Gregory Dubois-Felsmann: this is because the standard implementation of multi-object search involves uploading a temporary table to facilitate the spatial join, and this is not a feature available in Qserv in 2017. The SUIT team is currently implementing a workaround involving iteratively calling DAX dbserv one object at a time, so this capability will be available in the UI soon (Unknown User (xiuqin): 1-2 weeks).
Chris used the same approach - writing a script to loop over single-object DAX queries - to look at light curves for known objects.
Observed problems when trying to do queries covering modest fractions of a square degree - these seemed to fail "9 out of 10 times" but would sometimes succeed on retry or even on browser-back, browser-forward operations
NB: Object density in Stripe 82 is approximately 10^6 per square degree
Observed problems with the retrieval of postage stamps associated with individual observations: they tend to work fine in a fresh session, but fail in a variety of ways on retrying them or when changing the cutout size requested. In some cases requesting larger cutouts appeared to cause non-SDSS imagery to be displayed. Was comparing with the official SDSS DR13 data service as well as with Aladin.
The PDAC group all encouraged Chris to provide bug reports right away and not save them up. We are very interested in getting rapid feedback on problems observed.
Chris discussed the next set of things he would like to be able to do, following on from a discussion on Slack #dm-pdac yesterday.
(minutes-writing will continue later today)
imgserv and stack v13.x backward compatibility with Stripe 82 processing