--- On Sun, 12/2/07, Hiroshi Inoue <inoue@tpf.co.jp> wrote:

> Do you think it is an appropriate handling ?

This is a good question. I am not sure what an appropriate handling would be. My first test of '-infinity' (which resulted in a raised error) was with an ado recordset. Since you are questioning what is appropriate I thought that I should try a few different client front ends to see how they treated (+/-)infinity:

Client | -infinity | + infinity
ADO | <error> |9999-12-31 23:59:59|
Access/LT|1899-11-30 00:00:00|9999-12-31 23:59:59|
OOB(ODBC)|0001-01-03 00:00:00|9999-12-31 23:59:59|

ADO = ado recordset created in VB
Access/LT = a linked tabled in MS-Access (probably DAO/ODBC?)
OOB(ODBC) = Open Office.org Base ODBC connection to the Database.

Having done further tests, it seems that the error must be a bug with ADO since other clients have not a problem with making -infinity equal to the smallest time stamp that their respective data type can display. In my case, I do not really care about what time stamp values are displayed by the clients for (+/-) infinity since I use these values as an implied NULL value for time stamps. Unfortunately, ADO is broken and can't display any value for -infinity.

> How do you think the driver should treat -infinity ?

It seems that it already handles it correctly. I was mistaken in thinking that the translation of (+/-) infinity was handled by ODBC. )
Regards,
Richard Broersma Jr.

---------------------------(end of broadcast)---------------------------
TIP 6: explain analyze is your friend