* Experiments to run (Wednesday)
** Pin policy
*** 30s freshness, microcache, no jitter
*** 30s freshness, microcache, with jitter
*** Use whichever turns out better
** Microcaching
*** Same, with microcaching off
*** Whichever turns out to be better is our main result graph
** Freshness
*** 5, 15, (30), 60 sec freshness requirements
** Baseline
*** Rerun bypass-caching, either stock/modified Postgres



* Experiments to still run (Thursday)
** Higher points on 5, 15 freshness curves
   Start wherever the 60s curve started (relative to the peak)
** Main curve, no microcaching

* Paper bits
** DONE Varying freshness interval
** DONE Client library -> txcache/application library
** DONE State correct pin policy
** DONE Discuss MVCC related work
** DONE Write conclusion
** DONE Rewrite pincushion
** DONE Intro to DB section
** TODO DB section: how MVCC/snapshot isolation works
** DONE Sec 3: r/w transactions inherit the DB's isolation properties
** DONE Say more about truncating validity intervals in DB
** DONE Move validity interval example into text