PDA

View Full Version : Beyond relational databases



Argos
2005-May-24, 09:24 PM
An interesting article on future trends: There is more to data access than SQL (http://acmqueue.com/modules.php?name=Content&pa=showpage&pid=299)

mid
2005-May-25, 10:56 AM
You know, if people actually used RDBMS systems to do proper relational databases in the first place, I'd have a lot more sympathy for his arguments. But lets see if we can do those correctly before we try and move on.

These post-relational systems always hit the fundamental problem that they're slow as a jar of molasses immersed in liquid helium anyway.

Maksutov
2005-May-25, 11:23 AM
You know, if people actually used RDBMS systems to do proper relational databases in the first place, I'd have a lot more sympathy for his arguments. But lets see if we can do those correctly before we try and move on.

These post-relational systems always hit the fundamental problem that they're slow as a jar of molasses immersed in liquid helium anyway.
Right, plus you'd darn well better have ALL your relationships well defined and mapped out before attempting migration to such a thing. And from a functioning, debugged database that's 100% backed up.

It's not like you could just convert to delimited ASCII and convert.

Argos
2005-May-25, 02:01 PM
I think both of you have good points, but I can only agree with the author on that most of the times RDBMS´s use a cannon to shoot a fly. :)

jfribrg
2005-May-25, 02:09 PM
There is one area that he didn't mention that I would like to see. Extending the relational model to incorporate temporal aspects of data would be very useful. Much of the functionality of a financial database (and many others as well) deals with temporal data but the current relational databases do not explicitly support it. Temporal support is still a matter of each database application reinventing the wheel, much like data storage and retrieval in the old ISAM and VSAM systems. Temporal extensions to the relational model have been discussed in academia for years, but for some reason, these extensions have not been embraced in the real world.

mid
2005-May-25, 03:02 PM
most of the times RDBMS´s use a cannon to shoot a fly. :)

I'd definitely agree with that, but then it is often the way with computers. I've lost track of the number of times I've seen both shooting a fly with an ICBM (enterprise-class RDBMS systems to store a bunch of files, Adobe Photoshop being used for picture resizing, Maya to add a nice emboss effect to a logo etc.). But then I've also seen my fair share of seriously complex databases "running" (i.e. crawling) in Excel, which is like trying to transport an elephant across the Atlantic by sticking some waterwings on its legs.

aurora
2005-May-25, 04:37 PM
There is one area that he didn't mention that I would like to see. Extending the relational model to incorporate temporal aspects of data would be very useful. Much of the functionality of a financial database (and many others as well) deals with temporal data but the current relational databases do not explicitly support it. Temporal support is still a matter of each database application reinventing the wheel, much like data storage and retrieval in the old ISAM and VSAM systems. Temporal extensions to the relational model have been discussed in academia for years, but for some reason, these extensions have not been embraced in the real world.

One article that I read this week, purported to be describing the latest thought on the topic, went into a description of what RDBMS is not good for -- data sets that have real time collection of vast amounts of data (like telephone systems) and where the data will be infrequently (if ever) accessed. In that instance, the article suggested the extremely innovative solution of using a flat file(!).

Sometimes the old ways are the best ways.