In loosely coupled systems, such as systems connected via web services, is there any reason to attempt to design pessimistic locking systems with transactions? Or is it better to work on optimistic locking systems that provide multiple shared copies of a resource, and resolve conflicts with an effective collaboration system run by humans?

From a practical point of view, i’ve worked on systems where transactions and locking are attempted to be enforced on client-server SOAP transmissions. With all the delays and problems of the internet, deadlocking and corruption were common occurences, and it certainly makes sense to me that people trying to work around this problem with complex WS algorithms are certainly expending a lot of time and effort chasing an impractical goal. An interesting series of blog entries.

http://www.manageability.org/blog/stuff/whycantlockingbe/view http://www.manageability.org/blog/stuff/why-pessimistic-transactions-arent-practical http://seanmcgrath.blogspot.com/archives/20040502seanmcgratharchive.html#108391726581259860