On 7 Oct 2010, at 1:46am, Jay A. Kreibich wrote: > Much of the big "No-SQL" movement does exactly this. > The big secret is that most of these applications are still doing > relational data operations, they're just pushing the manipulations > (such as a JOIN) down into the client code, making the "database" a > more simple level storage engine. Many of the end results are the > same, however. It makes sense when performance is at an utmost > premium, and allows greater scalability, but it also increases > application complexity and tends to skirt some issues like > transactional safety. Those types of systems also trade flexibility > in the general case for complexity and speed in the common case.
Which in itself is a result of Moore's Law. SQL was invented back when DBMSs were needed because computers were slow and memory was expensive. ORDER BY was invented because of the incredible convenience of being able to retrieve your records in the order you wanted to process them in. Being able to create your records in one order and retrieve them in another was ridiculous luxury. Nowdays, although INDEXes are still important to save you from having to grovel through the whole file, sorting a thousand records doesn't require writing a temporary file. Read all the records from disk, then sort them in memory. Memory is cheap; processing is fast; who cares ? So nowdays it's possible to push some of those manipulations previously done at the database level to your application code. Which means you need less stuff in your DBMS. Which means it can be better optimised for general operations. Which is how you get No-SQL and similar unSQL things. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
