This has to do with the old debate of special purpose software versus general purpose.
If you NEVER want to use that data for ANYTHING else, or if you don't mind paying a lot to a team of developers to do ANYTHING you need out of that data, then the non-rdbms solution is perfectly valid.
If you want that data to be available to other sub-systems on short notice and without needing to engage developers to write a specialized interface, then a non-rdbms solution is totaly unadvised. A general purpose rdbms will be accessible with ease by ANY other team or special interest, without yet another major development effort being undertaken.
That debate was made many years ago, and the solutions have been available since then.
You chose a specialized - I believe the term in vogue nowadays is "bespoke" but I refuse to use it - data management storage architecture then you get a very good match for that special application and no match for anything else.
You want something more general that can be easily accessed and modified by any number of additional applications without additional development costs - not necessarily specified up front - then you use a general purpose data management product. You pay a hit on performance and you gain flexibility.
Horses for courses. -- Cheers Nuno Souto in sunny Sydney, Australia dbvision@xxxxxxxxxxxx Blake Wilson wrote,on my timestamp of 10/06/2011 1:22 AM:
However, the next release of one of our options will not have a RDBMS in the solution. It will be replaced by Apache Jackrabbit. The new system will have everything// treated as content, including grades, test questions and answers, discussion threads, syllabi, personal profiles, chat messages, and so on. This seems like quite a departure from normal RDBMS based solutions. Is this a good idea? Am I being a dinosaur by thinking that this is not a good idea? Do I need to keep up with the times? Is this the future of databases? This really looks to me like a return to design of 20 years ago.