On 9 February 2015 at 05:07, Matthew Hall <mh...@mhcomputing.net> wrote: > On Sun, Feb 08, 2015 at 11:07:59AM +0100, Michael Osipov wrote: >> How would you propose to use it? >> I would still need a random number or do you recomment to use a static one >> and simply increment during the lifetime of the application? >> >> If so, thing might cause trouble in the future because I would probably >> store the session information in a database for analysis issues and woudl >> have collisions. >> >> Michael > > Then you either have to use real big random's like UUID and accept it'd be > slow, or derive the numeric values from a native DB counter (for example, a > Redis counter, or a PostgreSQL sequence data type).
This is a separate issue from the pooling problem. Depending on how you are intending to create the sessions, it may be possible to use a cheaper method. For example combining system time with System.identityHashCode should work if only a single classloader is ever used at one time. > Matthew. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: user-unsubscr...@commons.apache.org > For additional commands, e-mail: user-h...@commons.apache.org > --------------------------------------------------------------------- To unsubscribe, e-mail: user-unsubscr...@commons.apache.org For additional commands, e-mail: user-h...@commons.apache.org