No notion of "accepted date" in hotcrp

[kmccurley]

We need an acceptance date for the URL to upload to publish.iacr.org. There is a submission date in hotcrp, but there doesn't seem to be a well defined 'acceptance date' (well anyway it doesn't appear to be recorded in the Paper table). There is an ActionLog in which changes are recorded to the decision, but I found at least two kinds:

Set decision: Conditional accept (papers 8, 86, 97, 193, 219, 221, 297, 335, 348, 355, 397, 447)

where the paperId is NULL, and individual ones:

Set decision: Conditional accept

where the paperId refers to an individual paper. Note that a paper can move between different decision values, and that some are ambiguous (e.g, "Conditional accept"). Moreover, we would have to search for the most record log item on a paper.

[kmccurley]

The best thing we might have is the date when reviews are due from the PC members. This can show up in several places in the Settings table of HotCRP. When I looked at tches2023_3, it had the following values in the Settings table:

+------------------+------------+
| name             | value      |
+------------------+------------+
| sub_open         | 1672693782 |
| rev_open         | 1672693932 |
| sub_sub          | 1673870399 |
| sub_reg          | 1673870399 |
| sub_update       | 1673870399 |
| pcrev_soft       | 1676375999 |
| pcrev_informtime | 1677150038 |
| pcrev_assigntime | 1677150038 |
| final_soft       | 1681559999 |
| __capability_gc  | 1687984074 |
+------------------+------------+

These are timestamps, and it looks like pcrev_soft is the "soft deadline" for reviews to be returned, which was 14 Feb 2023 11:59:59am UTC according to the form. Note that if there are subsequent rounds, they would show up after this date and I don't know what they would be called in the table (nothing in hotcrp is documented). The final_soft value is the deadline for submission of final versions, and I think the pcrev_informtime is when pc members are last informed that they are assigned a review.

[kmccurley]

Trying to reverse engineer the individual acceptance date from the ActionLog seems very fragile, in part because entries can look like "Conditional accept" or "Merged accept" or "Accept as short paper" depending on whether the chair has configured other kinds of accept decisions. Moreover we see things like

Set decision: Accepted (papers 4, 6, 19, 25, 29, 31, 38, 54, 60, 64, 67, 72, 73, 75, 79, 92, 96, 97, 98, 99, 101, 102, 103, 104, 109, 111, 114, 128, 131, 136, 137, 152, 161, 165, 167, 176, 178, 182, 184, 185, 186, 188, 201, 202, 206, 208, 218, 227, 230, 234, 235, 241, 250, 257, 262, 265, 267, 272, 298, 299, 301, 302, 304, 307, 313, 314, 323, 324, 331, 336, 342, 350, 352, 356, 365, 369, 372, 373, 374, 376, 378, 379, 380, 386, 388, 390, 393, 396, 399, 403, 410, 411, 418, 419, 426, 429, 431, 440, 445, 448, 452, 454, 461, 464, 481, 482, 484, 489, 494)

As an approximate solution, I suggest using

select max(timestamp)  from ActionLog where action like 'Set decision:%' and action like '%ccept%';

This gives the timestamp of the last log event for acceptance among all papers. It's at least a safe upper bound.

[kmccurley]

Fixed with https://github.com/IACR/hotcrp/commit/763eb88a809da386a93c909e090d3fb483925eae