Re: How to make RDBMS ?
Date: Thu, 10 Jan 2002 18:10:39 +0000 (UTC)
:>Todd, :>> I have lately come to believe that this is a misconception. What is :>> undeniably true is that the OS problem was attacked first, but I think the :>> DB problem is inherently simpler. :> :>I had actually taught the internals of a much earlier OS, and I also think :>the complexity is of the same order. :>
: That might have been true in the past but certainly not today. Optimizer alone
: has so many areas that the industry implemented only a tiny bit of the
And operating systems are immune to optimization? Wake up. Kernel hackers eat, sleep, breathe, and piss optimization.
: research available. Database features are coming with the speed that
: makes very hard for ordinary database application developer to keep up.
OS designers are very familiar with application developer lag. POSIX threads were standardized a decade ago (IIRC 92 or 88), the standard was everywhere four years ago, and only now is threading considered as a base block in many application designs. This sort of thing happens everywhere.
: User programmed domains, extensible indexing, user-defined statistics to
: name just a few oracle features. (I didn't meet any index developer
: yet, however;-).
These are all things that are important, and have analogies with OSes. Also, user defined domains should have been in our hands 10 years ago. The delay has been absurd.
: In short, there is so much database does already, and will do in the future,
: whereas operating system area is pretty much frozen.
That is an almost totally religious statement, and a very misinformed one. Your statement is a classic example of focusing on one area of research to the exclusion of basic awareness of others. You are somewhat exonerated by the Microsoft monopoly obstructing progress from reaching the consumer, but still...OS research has picked up speed recently, and some areas I try to follow are advanced filesystems (which are becoming more like databases every day), the unstoppable advance of graphics, capabilities, and ways of structuring a kernel for stability & performance (-- which may be more generally said to be able to warp system behavior towards the needs of applications much as LISP programs warp the language to suit themselves). Other systems are working on componentizing everything in an attempt to destory the current concept of application.
Just look at the last several years at what has emerged into consumer OSes. Threads? Remember when 'Z-buffer' wasn't a word? TCP/IP everywhere? (...and might I add, rewritten everywhere to handle the load)
If none of this has bothered you, Microsoft should. They have pointed out that there is really nothing in the definition of operating system that prevents it from encompassing everything on the system. From a technical point, everything a database can do can be implemented as a core part of the OS. Received on Thu Jan 10 2002 - 19:10:39 CET