160 likes | 172 Views
Learn about component standards like CORBA, DCOM/ActiveX, JavaBeans, COM+, and more for a generalized framework. Find insights on pros and cons, usage scenarios, and future outlook.
E N D
ATLAS and Component Standards Lassi A. Tuura, CERN
Terminology & Background • Component Architecture Framework • One framework for reconstruction and analysiscode must be easily moved between the two • Data-centric problem view: • Data objects physicist tend to think about • Calculation objects transform and operate on dataand are external to the data • Transient intermediate data objects Lassi A. Tuura, CERN
Standards We Will Use • Object Database Standard (ODMG) Lassi A. Tuura, CERN
Other Standards • CORBA • DCOM/ActiveX • Java Beans • COM+ Lassi A. Tuura, CERN
CORBA: pros • Relatively stable standard, free brokers exist • Interoperable across languages and platforms • Reasonably sound from OO point of view • Good for client-server apps Lassi A. Tuura, CERN
CORBA: cons • A component architecture, not a framework • Incomprehensible to a casual developer • Speed versus encapsulation issues with small components • With an OODBMS, what does it really by us? • Inadequacy of free implementations versus cost of commercial ones? Lassi A. Tuura, CERN
DCOM/ActiveX: pros • Widely used • Sort of interoperable across languages • Good for client-server apps Lassi A. Tuura, CERN
DCOM/ActiveX: cons • Standard still in fluxespecially ActiveX • Proprietary, limited platform availability • Hideously complicated, well-formed code resembles line noise • Lacking OO support • Like CORBA: only a component architecture, performance limitations, not necessarily well suited to the problem domain Lassi A. Tuura, CERN
JavaBeans: pros • Nice, clean and simple design • Gateways to CORBA and COM exist • Platform independent GUI development comes for free • A promising candidate for the future Lassi A. Tuura, CERN
JavaBeans: cons • Close to being a framework, but still not thereprobably quickly done • Platform independentbut simple only in a pure Java environment • Immaturity and instability of the Java development tools • Unstable and proprietary specification Lassi A. Tuura, CERN
COM+ • Don’t know much yetnot yet available, estimated in large scale use by end of 1999 • Addresses COM’s complexity and OO deficiency problems • Component model approximately as simple as JavaBeans • Simplifications made by changing the C++ language! Lassi A. Tuura, CERN
Alternative Non-Standards • IRIS Explorer • AVS/Express • ROOT Lassi A. Tuura, CERN
IRIS Explorer • Commercial, proprietary component interface • High risk • Not object oriented • Abysmal bang for the buck • Technical deficiencies • Unsuitability to non-visual batch execution? • Virtually all parts of interest to us seem to be available for free or very low cost • Leaves important infrastructure gaps Lassi A. Tuura, CERN
AVS/Express • Nice object and component models • Nice to use • Too expensive • Object model conflicts with OODBMS? • Few components useful to uswould have to develop many more ourselves (seems to be less effort than with IRIS Explorer, however) • Like IRIS Explorer: proprietary, high risk, leaves gaps in the framework Lassi A. Tuura, CERN
ROOT • BAD design Lassi A. Tuura, CERN
Conclusions • General framework with maximal simplicity • Use CORBA, JavaBeans, DCE in areas less exposed to physicist developers: • Online (details: Bob Jones) • User interfaces • Graphics Lassi A. Tuura, CERN