Re: ? SET ISOLATION

At 02:58 AM 3/9/95 GMT, W Mat Waites wrote: }In article <3jiggo$o69@cssun.mathcs.emory.edu> xrizzi@arcride.edu.ar writes: }[ some stuff deleted ] }> }> The locks level of the table used in the CURSOR is to ROW. }> }> From different processes is NOT possible to agree to rows NOT selected }>by the CURSOR FOR UPDATE. }> }> The messages error of informix is the following: }> }> - 246 could not I give an indexed read to get next row. }> }> Please, us interesting to know if the explained procedure is incomplete }>or exists some another alternative that we have not considered. } }It sounds like you are hitting the "adjacent row key locking" }problem. This is supposedly fixed in 6.0. } }When a row is locked, the row before it and after it is also locked. Actually, adjacent key locking, should only lock the next key, not the previous. This was known as the "twin problem". In the DSA releases (6.0 and beyond), the locking scheme was changed, to 'key value locking' which only locks the current record. In order to do this, the index structure was changed slightly, adding an extra byte per index row. For this reason, when migrating from 5.x to 6 or 7, all indexes must be converted (drop and rebuild). Jon } }Kooky, eh. You must either spend alot of time working out a way }to get around this problem, or go for 6.0/7.1. } }Good Luck, } }Mat }-- }W Mat Waites | mat@rencorp.com }REN Corporation - USA | (615) 353-4110 } } ================================================================== Jon Vemo Internet: jvemo@cyberspace.com Bothell, WA USA ------------------------------------------------------------------ Watch out for road-kill on the information superhighway....SPLAT!! ==================================================================