On Oct 3, 2012, at 7:53 AM, Paul Homer <[email protected]> wrote: > If instead, programmers just built little pieces, and it was the computer > itself that was responsible for assembling it all together into mega-systems, > then we could reach scales that are unimaginable today. To do this of course, > the pieces would have to be tightly organized.
The missing element is an algorithmic method for decomposing the representation of large, distributed systems that is general enough that it does not imply significant restrictions on operator implementation at the level of individual piece such that the programmer has to be aware of the whole system implementation. The details of the local implementation will matter much less but it will also be a very different type of interface than programmers are used to designing toward. Programming environments improperly conflate selecting data models and operators with selecting data structures and algorithms. Reimplementation is common because I need sorting, maps, vectors, etc in the abstract to build the software but the algorithms someone else may use to optimally implement them for their data model is pathological for my data model. The algorithms used to operate on a data model are not separable from the data model but most programming environments treat them as though they are. Large systems are not nicely decomposable because they impose global implementation details beyond the interfaces of the pieces. A canonical example is the decomposition of an ad hoc relational join operation. Common data model representations impose the use of algorithms that are pathological for this purpose. You can't get there, or even close, with the data structures commonly used to represent data models. At the root, it is a data structure problem. -- J. Andrew Rogers _______________________________________________ fonc mailing list [email protected] http://vpri.org/mailman/listinfo/fonc
