Web Services and Transactions
May 7, 2004
This article sucks on so many different levels, I don’t feel like pointing them out in detail. The short version is: Mentioning XA in an SOA context is a dumb idea, and I consider WSIF to be fundamentally flawed.
About
This page contains a single entry from Stefan Tilkov's Random Stuff posted on May 7, 2004 4:37 PM. The previous post in this blog was XML for Data Modeling?. The next post in this blog is New Features in EJB 3. Many more can be found on the main index page or by looking through the archives.
Comments
Hello Stefan, Can you please elaborate a little bit more on why you believe WSIF is flawed Thanks
Posted by: Sergey Beryozkin at May 10, 2004 1:33 PM | link
I believe the whole idea suggests that whether you use RMI/IIOP, HTTP, JMS or any other means of communication is just a low-level communication detail. I don’t think it is, and I think that standardizing on the programming language API level as opposed to the transport/communication level is a bad idea.
Posted by: Stefan Tilkov at May 13, 2004 9:26 AM | link
I agree - it sucks. I have never spoken with this lady who seems to have been listening to my presentation with less than her full attention.
I did mention XA in the context that in an SOA each component (executable) needs to be responsible for it’s unit of work so that it becomes an atomic unit. I stand by that because I need to co-ordinate MQ and local databases for each component of the service to make it easier to recover from failure. This was stated in an MQ context because many customers fail to realise that they issue MQGET, SQL Update, SQL Commit and then fail. The MQGET gets rolled back unless you have XA co-ordination enabled. The restart then processes the message again - voilla - a duplicate.
I apologize for the article.
Posted by: Peter Rhys Jenkins at June 3, 2004 9:31 PM | link