[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: transaction scoping
I see your points. More generically speaking, my question could have been:
how do I write methods to be transactionally safe on their own, as well as
when they're called as part of a larger logical transaction. My follow-up
therefor:
I was a developer on the IBM San Francisco project, which resembles some of
the elements of ozone (factory, server-side objects accessed by proxy) and I
used the framework in several application development projects. One of the
main problems we encountered was the need for transactional systems to bunch
up transactions. Think about batch transactions. You have a 100 objects that
you want to update in batch, and you want the whole batch to succeed or
fail. If something goes wrong in processing objext #25, you don't want the
other 75 to possibly fail as well, so you just roll the whole batch back. We
had to write our own transaction wrapper to be able to do this. IMHO, it
would be a really good thing if a project such as ozone would solve this
problem early on.
cheers
Rob