Comments on: Two kinds of DBMS extensibility http://www.dbms2.com/2005/12/12/two-kinds-of-dbms-extensibility/ Choices in data management and analysis Thu, 09 Feb 2012 09:22:14 +0000 hourly 1 http://wordpress.org/?v=3.0.3 By: Curt Monash http://www.dbms2.com/2005/12/12/two-kinds-of-dbms-extensibility/#comment-194 Curt Monash Thu, 15 Dec 2005 18:23:34 +0000 http://www.dbms2.com/?p=37#comment-194 Logically there may not be. But standard DBMS access methods assumes that values are sortable in a single dimension (numeric or alphabetic). If you tried to use operators that ignore the sort order, performance in the UDT implementation could be inconceivably bad, often in table-scan range, or even repeated-table-scan. Try to do geographical operations of "nearness" in a single column without a geographic datatype. Oracle tried ("hhencode"). The performance wasn't pretty. How would you do a text search without scanning every row in the column? The mind boggles at the processing required (except maybe if you can keep the whole database in RAM). Logically there may not be. But standard DBMS access methods assumes that values are sortable in a single dimension (numeric or alphabetic). If you tried to use operators that ignore the sort order, performance in the UDT implementation could be inconceivably bad, often in table-scan range, or even repeated-table-scan.

Try to do geographical operations of “nearness” in a single column without a geographic datatype. Oracle tried (“hhencode”). The performance wasn’t pretty. How would you do a text search without scanning every row in the column? The mind boggles at the processing required (except maybe if you can keep the whole database in RAM).

]]> By: Eric http://www.dbms2.com/2005/12/12/two-kinds-of-dbms-extensibility/#comment-192 Eric Thu, 15 Dec 2005 12:39:41 +0000 http://www.dbms2.com/?p=37#comment-192 While I understand that these things have been implemented differently, I'm unable to detect any fundamental difference between these two... While I understand that these things have been implemented differently, I’m unable to detect any fundamental difference between these two…

]]>