Saturday, December 29, 2007
Transaction Models, Again
Unfortunately, blogs can sometimes lose the nice nested associations of discussions that characterized Usenet, but they remain an interesting and more accessible way to debate ideas. Mark replied back with a reply to my previous comment on transaction futures with a clarification. What I meant to say was that I more or less explicitly disagree with the position as he's clarified it: after looking at attempts to reconcile multi-party choreographies and efforts around BTP, I have come to believe that large scale systems that require a global coordinator are destined to fail. The future belongs to a different paradigm, where bilateral negotiation and local enforcement of "system" invariants are likely to be building blocks for successful applications. I am confident that infrastructure and frameworks will play an in important role, but in this case, I am inclined to paraphrase Nietzsche: "The Coordinator is dead!"