Comments on: Relational Databases Considered Harmful http://www.airs.com/blog/archives/171 Ian Lance Taylor Sun, 07 Mar 2010 17:12:02 +0000 http://wordpress.org/?v=2.9.1 hourly 1 By: eternaleye http://www.airs.com/blog/archives/171/comment-page-1#comment-18289 eternaleye Sun, 07 Mar 2010 17:12:02 +0000 http://www.airs.com/blog/archives/171#comment-18289 Came across this while reading through your archives after finding your site via your excellent linker internals posts. I think that in this view of "how databases should be" you would rather like the SPARQL/RDF system that is becoming popular for the "semantic {web,desktop,etc.}". Check out Nepomuk in KDE 4 and the Virtuoso RDF database engine/server that it's built on these days. Came across this while reading through your archives after finding your site via your excellent linker internals posts. I think that in this view of “how databases should be” you would rather like the SPARQL/RDF system that is becoming popular for the “semantic {web,desktop,etc.}”. Check out Nepomuk in KDE 4 and the Virtuoso RDF database engine/server that it’s built on these days.

]]> By: Ian Lance Taylor http://www.airs.com/blog/archives/171/comment-page-1#comment-12185 Ian Lance Taylor Sat, 05 Apr 2008 01:26:45 +0000 http://www.airs.com/blog/archives/171#comment-12185 Exactly. Thanks for the note. Exactly. Thanks for the note.

]]> By: rskrishnan http://www.airs.com/blog/archives/171/comment-page-1#comment-12174 rskrishnan Fri, 04 Apr 2008 20:25:30 +0000 http://www.airs.com/blog/archives/171#comment-12174 Could'nt agree more - i.e. relational databases suck - with no doubt whatsoever ... at least in my mind. It starts out very nicely - with a "logical" organization to the data schema, but pretty soon you land up with a rats nest of indices, and "flattened tables" which all need to be kept in sync. So the problem goes from "hey here's a black box to put my data into" ... to ... something like ... "if you're loading a LOT of data, then turn off index consistency checks, since they are slow - then load data, then update your indices", or ... "if your report is large go to flattened_table_xyz - not to the clean structure, and by the way we update the flattened table only once a week" !! More interestingly - I don't think the rdbms ever delivered on it's promise of "your data all the time, anytime". Every reasonably large company that I know of spends many thousands of $$ on buying servers, disks, networks, and software -- all to run a measly GL/AR/AP accounting system ... and then guard access to the system like it was their first born! Could’nt agree more – i.e. relational databases suck – with no doubt whatsoever … at least in my mind.

It starts out very nicely – with a “logical” organization to the data schema, but pretty soon you land up with a rats nest of indices, and “flattened tables” which all need to be kept in sync. So the problem goes from “hey here’s a black box to put my data into” … to … something like … “if you’re loading a LOT of data, then turn off index consistency checks, since they are slow – then load data, then update your indices”, or … “if your report is large go to flattened_table_xyz – not to the clean structure, and by the way we update the flattened table only once a week” !!

More interestingly – I don’t think the rdbms ever delivered on it’s promise of “your data all the time, anytime”.

Every reasonably large company that I know of spends many thousands of $$ on buying servers, disks, networks, and software — all to run a measly GL/AR/AP accounting system … and then guard access to the system like it was their first born!

]]>