One of our applications is an energy trading platform, parts of which have a stringent bug-fix time requirement: 15 minutes from bug report till fix or our customer gets hit with crippling fines. Those applications are deployed without sealing them, so when we get a bug report we connect to the machine and see an open debugger on the thread that failed while the rest of the application happily purs along; we modify the code and/or data in the debugger and hit the continue button, nearly always within this 15 min time frame. We then code up a permanent fix in our shop and hoist this permanent patch in the production server /while/ it is running. Again something that is mostly impossible in dead-code systems.
Mr. Reinout Heeck is talking about Smalltalk. Recently it has come in the news again because this company repurposed their Smalltalk VM to run Ruby, with great performance boosts. Their VM is not entirely ready but is meant to replace most of the Ruby on Rails stack. They're calling it MagLev (geddit? rails-maglev hee hee :)).
It sure sounds mightily impressive: an environment that automatically persists objects, has them available to many concurrent worker threads (including across hosts) with a fully transactional model and is zippy to boost... now almost available for a scripting language that has its users cooing with pleasure! Throw in some of that live debugger salsa and I'm sold.
Sometimes, life is good. Cue the Louis Armstrong playlist, please.
No comments:
Post a Comment