Re: [Fwd: Re: a trigger question]

rkusenet wrote: > "Jonathan Leffler" <jleffler@earthlink.net> wrote >>I'm not sure that I can think of a way to reconcile a successful >>insert with doubling operation and a failure to respond to an >>exception -- it makes me think of creepy-crawlies. Somewhere along >>the line, someone - possibly you, Ravi - suggested that this should be >>the expected behaviour. Does anybody have a reference to the manuals >>where such niceties are discussed (page number and edition would be >>helpful)? > > I don't remember about manual, but I think the following is considered > as expected behaviour in an unlogged database:- > >[...] > > Unlogged database makes serious compromise in data integrity. > I don't know about other databases, but in Informix, in an unlogged > database, it is very easy to run into a situation when a NOT NULL > column can have NULL values. Most other DBMS do not have a concept equivalent to Informix's unlogged database. After various investigations, not excluding loooking at a presentation of mine from February this year, Ravi's correct in terms of what happens, and it is documented as what happens. It is a good reason for always using a logged database with IDS; in fact, I can't think of many good reasons for using an unlogged database in IDS (the main one is that you need to interwork with an unlogged database where you cannot get the supplier of that database to switch to a logged database). -- Jonathan Leffler #include <disclaimer.h> Email: jleffler@earthlink.net, jleffler@us.ibm.com Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/