On Wed, Mar 18, 2009 at 1:21 PM, Brian Aker <[email protected]> wrote:
> Hi! > > The option "sysdate-is-now"... what should this be? It is basically there > for Oracle compatibility, but it does give you the advantage of a time > creation that occurs in row. > > The option "interactive_timeout". I've found over the years that users will > be confused about this and the wait timeout. What should the functionality > be? Should we let clients override the system maximums? > "sysdate-is-now" is there so that folks who use sysdate won't get into trouble with replication. See http://www.pythian.com/news/1295/does-anybody-really-know-what-time-it-is I believe the answer is "replication should use deterministic functions as often as possible." If it were up to me, sysdate() would be an alias to current_timestamp(), and it should be difficult to use a non-deterministic function in replication. As for interactive_timeout -- while usually, yes, wait_timeout is usually more the culprit, it's useful to have 2 separate variables for the 2 separate concepts: 1) How long to allow a thread to sleep before killing it in a non-interactive session 2) How long to allow a thread to sleep before killing it in an interactive session I'd vote for "keep both, but change wait_timeout to noninteractive_timeout to be more clear." The confusion I've seen is that there's more than one parameter, and they're not usually cross-referenced. If documentation that talks about interactive_timeout says "see also noninteractive_timeout" and vice versa, it would be less confusing. However, having both parameters is very useful. -- - Sheeri K. Cabral http://tinyurl.com/mysqlbook will take you to the Amazon.com page for my upcoming book, "MySQL Administrator's Bible".
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
