Java creator James Gosling’s remarks to the NIST group examining the requirements for real-time extensions to Java (CI No 3,545) prompted more than one member of the group, including Mitre Corp’s Douglas Jensen to respond to Gosling’s thoughts on the use of asynchronous events. Jensen notes that Gosling hasn’t been following the group’s discussion too closely, and believes that when he [Gosling] has time to read the messages [on asynchronous exceptions] he will see that we are not advocating a Unix-like exception mechanism. Jensen says the group has stated that signals are not a good asynchronous event model, and such mechanisms deserve to be deprecated in Java. Jensen says that in his experience coding defensively (a term that implies an inappropriately negative mind-set about asynchronous events) can be quite natural. He says: perhaps a good analogy is to non- determinism, in that it can be quite effective to embrace and exploit it rather than fight it. Then the coding does not devolve to the equivalent of polling, which is one of things asynchronous events strive to avoid. Furthermore, he says James’ assertion that ‘resume’ (continuation) semantics are equivalent to a separate high priority thread waiting (polling) for ‘exception events’ is simply incorrect in the Alpha distributed (trans-node) thread style programming model which needs the comprehensive type of asynchronous event facility I advocate. Such a separate thread would literally break the programming model. A comprehensive and carefully thought out asynchronous event facility runs counter to convention wisdom. But conventional wisdom comes from people who have never been in the business of writing distributed software (in our technical sense – e.g., maximizing the accrued utility of time-constrained thread activities) real-time programs which must adapt to unknowable dynamic changes in the execution environment while providing dependable behavior. Jensen finds it ironic that so much intellectual activity in this [Java real-time] group is expended on trying to reconcile the inherent dynamic personality of Java with the traditional static style of centralized synchronous real-time – when Java’s dynamic personality is such a natural fit with the dynamic style required for distributed asynchronous (=realistic) real-time systems. Why Gosling ever got involved with real-time Java issues in the first place is difficult to fathom given this is not his traditional field of expertise. He’s also got a million other things to do at the moment, not least of which is testifying before the US Government’s antitrust trial against Microsoft Corp in Washington DC. The uncharitable view would be that Sun has pitched Java creator Gosling into the real-time Java debate to give its case more weight against the splinter group forming around Hewlett-Packard Co and Microsoft. HP and Sun are both using the National Institute of Standards and Technology group work as the starting point for their respective real-time Java developments.
