What are you going to base your concurrency work on? I wonder because I skimmed through Marcel Kornacker's theses, and for example it mentioned use of predicate locking (among other interesting things). Do you have any ideas yet on how to handle that?
I've read a paper[1] on concurrency and recovery by Kornacker, C. Mohan and Hellerstein, which is presumably pretty close to Kornacker's thesis work (which I'll definitely read later). My understanding is that there's no need to implement predicate locking for GiST any more than for B+-trees -- it's a nice feature to prevent phantoms, but not a requirement for implementing page-level concurrency. So I haven't given any thought to implementing it. If I've misunderstood, please let me know.
-Neil
[1] http://db.cs.berkeley.edu/papers/sigmod97-gist.pdf
---------------------------(end of broadcast)--------------------------- TIP 2: you can get off all lists at once with the unregister command (send "unregister YourEmailAddressHere" to [EMAIL PROTECTED])
