- From: <keshlam@us.ibm.com>
- Date: Mon, 5 Jun 2000 15:49:21 -0400
- To: "xml-uri@w3.org" <xml-uri@w3.org>
>IMHO the cycles you spend on this are completely dwarfed by the cycles >you spend on general-purpose XML parsing anyway. Possibly. I don't have a strong answer, just a question. >As is said in _Software Elements of Style_: A program that uses matrixes >smaller than 10 x 10 can afford to neglect the time it takes to initialize >them. If the matrixes are larger than 10 x 10, it can also afford to >neglect initialization time --- because computation time will swamp it. That's true for tasks that run once. It's much less true for tasks that run frequently, as on a heavily-loaded server. Initialization time adds up when you start talking about many thousands of invocations per day, and may be the difference between a solution that you can afford to roll out and one you can't. (Or between technology you can sell to your customers and the alternative your competetor is marketing.) Ours is the only industry that habitually considers anything less than an order of magnitude difference to be insignificant. Which is one reason our gigabye/gigacycle PCs don't feel all that much larger or faster. I've no objection to XML's decision to trade off some efficiency in favor of other priorities. But when it's possible to do something efficiently without excessive impact on other issues... ______________________________________ Joe Kesselman / IBM Research
Received on Monday, 5 June 2000 15:49:38 UTC