Friday, May 12, 2006

Oracle ANSI join security bug. Bad news again.

Trying to figure out a workaround, I have found that these DMLs that break authorization, also can not be AUDIT TRAILED.

Thursday, May 04, 2006

Followup: Oracle block size limit on different platforms

There was a long and annoying dispute on one of the forums about why a maximum block size on Windows (and Linux) platform is 16K, while on "serious" platforms like HP, and Sun, it is 32K. I had a wild suggestion, a bold guess, that the limit has been set based on some business strategy, that should push users to migrate from "minor" platforms towards "serious" ones. My assumption was (and still is) based on a fact, that couple of years ago Oracle had different prices for Intel and RISC CPUs, and having a 32Kb block size could be one more reason for users to migrate on more expensive RISC platforms.I never believed in arguments of other people that the 16K limitation had been caused by some technical deficiencies of "minor" platforms. Sure, it is clear that in Windows, in Linux and even in DOS, it is possible to do I/O by 32Kb chunks. However, nobody could neither confirm nor disprove that the idea of 16Kb limit was a business driven decision. Bill S, has kindly contacted Tom Kyte and the OakTable members. But, to my understanding, nobody could remember the reasoning ( So, I have decided to contact Oracle Support. I have created a "Request for Improvement". My idea was: Support people may point out for me a technical incapability of "minor" platforms, if there is such.Oracle had accepted this RFI and registered as a Bug(!) #5196815. Funny, they call requests for improvements "bugs". May be because people are bugging them by their requests? :)))

However, they have not answered why it had been decided to have 16K on Windows and Linux. That indirectly confirms my suggestion about a "business driven decision".