I generally browse some of the techie blogs from blogs.oracle.com or dzone.com. Since am yet not much into blogging ,am still on lookout for the new and interesting blogs. Any suggestions to the same would be welcome.

Speaking of something interesting, came across this piece of nicely written blog from Jonathan Gennick. I wasnt aware that a query which had been running fine in versions before 10g, might suddenly fail, with far-reaching impact to some of the apps.

As I read through the article, I found it really intriguing and scary. As a matter of fact , it seems that given the new built in intelligence of the much publicised "CBO" from oracle, migrating from other versions to this new version would need a rigorous amount of testing maybe with the live data set and analysing the tables (as in this case).

Dont get me wrong. I am not against CBO. But sometimes it seems to be overdoing things.
I have been working on 10g for sometime and my experiences with it have been mixed. Had a near fatal experienceonce thanks to 10g. I wasnt aware that 10gR1 doesnt like the index rebuild to go online while the data is being loaded in direct path mode. It corrupted the whole damn index!!!!!!! Thankfully the system was still in trial.

And speaking of CBO, its way too dynamic and would certainly not recommend it if you are dealing with VLDB. My experiences with it havent been too great. Oracle certainly needs to take new steps in parallelism, if it wants scalable ETL tools to be built using PL/SQL or OWB. They have moved forward, no doubt, but yet some of the basic parallelism is still tough to implement. Take the simple case of unloading data from two tables going in parallel. Most of the ETL tools would run two threads and unload the two tables in parallel, almost by default. But I am not sure if thats the case with OWB . Pl/SQL makes it tougher to implement. It canbe done, but there isnt a simple way to do this yet. (Any ideas/suggestions welcome).