> Although i am not aware of the roots of this discussion but would like
> to
> comment at this point .
>
> When we work with sequences an aborted transaction does have
> a permanent effect on the last value of sequence. Is this behaviour
> not a violation of above defination of transaction ?
I believe you are correct, which is probably why Oracle offers
serialized sequences for those who want them to be. Sequences have been
explicitly documented as functioning the way they do as a concession for
performance.
