As many of you know, I've been working sparingly on Prevayler during the past couple of years, and after seeing this thread popping up at TSS, I had to find some time to post this. Chill, this is not going to be yet another zealot versus hater kinda thing.
In fact, I want to apologize for having contributed to the enormous waste of time that were all the discussions of whether Prevayler was or not scaleable, suitable, better, faster, lighter, whatever. Mike Spille is damn right in his technical criticism, Prevayler is nothing but a clever log'n'snapshot mechanism you could use in some limited number of cases, and even then, there are some nasty pitfalls we tried to fix without much success, and ones we just couldn't do anything about, so you still need to be aware of them. And that's it. Prevayler is not a database replacement, and it's not spiritual salvation or eternal corruption, either. It's just a clean and tested implementation of a log'n'snapshot mechanism that should be used when it's convenient and safe to. Somehow I failed, when documenting it, to make it very clear what those situations are, and this has caused much, much trouble.
Actually, Cameron described it better than any of us involved in the project did, which is kind of embarassing: "It's just a way to keep some of your live Java objects backed up in a file on disk for the next time you restart your application."
I'm thankful for it being a fun project that got me started on the whole opensource Java thing, got me writing articles, got me in touch with agile software development methodologies, and offered me the opportunity to learn a lot while working on it. And, again, that's it. Now please, let's just get over this whole thing.
Previous Topic
