Re: lock modes, isolation levels, and scanning

Jacob Salomon wrote: > June Tong responded: > > You just need a little help. Check out my unofficial FAQ: > > http://www.geocities.com/SiliconValley/Bridge/4578/faq.html > > In the OnLine section (hmm, I wonder if this is the right place for > > it), there is the question: > > "If I declare a cursor using an isolation higher than dirty read, can > > I get it to skip locked rows? It seems like when I hit a lock and > > get an error, the next FETCH will return the same error." > > > > I think this will answer your question. > > Hi June, > > I just took a gander (not to be confused with ye olde undomesticated > semi-aquatic avian) at your solution. Loved it! > > I mentioned it to my friend Mike at work here. He asked a simple > question: WHY? Why would you want to read under these circumstances? > If your query requires correct data (hence, the need for the committed > read) then you have sacrificed the accuracy of your query's active set > by skipping the locked rows. You certainly wouldn't want to sum on > these rows! > > On the other hand, if that kind of pinpoint accuracy is not needed, you > might as well have used dirty read in the first place. Hi, Jake, how's life? I have no idea why. I have all the same questions and concerns that you have. (Maybe I should put them in the FAQ.) But people keep asking (that's why it's a FAQ), and that's the FA (friendly answer :-). June -- june_t@hotmail.com Grounded in Palo Alto, living on M&M's (plain)