That won't work, though. If I have a single-threaded system, relying on blocking function calls (e.g., for database access), then I can't check the clock during those calls.Another way to implement it would be for the thread to check every X seconds or milliseconds if the latency time has run out or not.
Are there potentially other problems for a single-threaded server?Not that I've noticed.
I'm not sure, but I believe making the bounded-latency features optional would meet this requirement.BTW, the CAP requirements draft does mention these features:
"Support functionality such that the client is not forced to remain connected and waiting while a command is in progress. One possible way for the protocol to meet this requirement would be to allow the CU to abort a command that is taking too long."
-- /==============================================================\ |John Stracke | http://www.ecal.com |My opinions are my own.| |Chief Scientist |=============================================| |eCal Corp. |Never mind the GUIs--Unix won't be for the | |francis@xxxxxxxx|masses until we fix backspace & delete. | \==============================================================/