tag:blogger.com,1999:blog-11463104.post111842191485223109..comments2021-10-19T01:56:24.241-07:00Comments on Robert G. Freeman's Blog: ORA_ROWSCN.... New 10g pseudo ColumnRobert Freemanhttp://www.blogger.com/profile/11021657343453201495noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-11463104.post-1118531498450227362005-06-11T16:11:00.000-07:002005-06-11T16:11:00.000-07:00This feature certainly does look interesting when ...This feature certainly does look interesting when I do get around to and having time for experimenting with html db I will try to remember this part. Maybe I have been lazy on the html db side (not really just too busy) but 2 things on my "waiting for list" include 10g release 2 and the tusc html db book. Thanks for noting this.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-11463104.post-1118513306769373342005-06-11T11:08:00.000-07:002005-06-11T11:08:00.000-07:00Great thoughts Tom!! Thanks for coming!Great thoughts Tom!! Thanks for coming!Robert Freemanhttps://www.blogger.com/profile/11021657343453201495noreply@blogger.comtag:blogger.com,1999:blog-11463104.post-1118434906372186362005-06-10T13:21:00.000-07:002005-06-10T13:21:00.000-07:00Robert -- ora_rowscn is cool -- but two things.fir...Robert -- <BR/><BR/>ora_rowscn is cool -- but two things.<BR/><BR/>first, you cannot (unfortunately) ALTER it "on", the table must be rebuilt. The reason why the table must be rebuilt is the second thing:<BR/><BR/>second, it adds 6 bytes of "overhead" per row. A little smaller than a date/timestamp and maybe a little bigger than a number.<BR/><BR/>BUT, you can turn the ora_rowscn into a timestamp using a function to map scns to time, showing the last modification of a row (for a while, as long as the SCN can be mapped) and it is totally non-intrusive to existing applications. <BR/><BR/>That last point makes it a good choice for apps that need to do this but coexist with other "not so well built apps" that do a select * and would be really upset if a new timestamp column appears.<BR/><BR/>Also, no triggers to write to maintain timestamps and such.<BR/><BR/>HTML DB does a md5 checksum of all of the columns read out and that can be a bit of a CPU user (the md5 algorithm) so if you actually find this easier (using the wizard and letting it do it for me was pretty easy), you might find this also to consume slightly less CPU as you are not computing that md5 checksum twice.Thomas Kytehttps://www.blogger.com/profile/00933377956793659534noreply@blogger.com