Actually, the one thing that bothered me in this whole business was that we were doing fake xml all along. I played with it assuming we're just using the xml-like text to find various ways of building huge strings, which is much more of a life-like situation.

I know I'd never try to write real XML walking (umm... doesn't translate well, take it as "in a pedestrian way"). I'd rather use cursortoxml() in several levels, if need be, or a XMLDOM object or anything, just like you said. Note that the fastest solution in my benchmarks (HBrennan) doesn't even reduce spaces between tags - a c(120) field would have all of the 120 spaces between the tags, which is probably OK for a XML parser, but is surely not nice.

However, this whole game did have a purpose, even though it may not have been clear in the outset :). We've found a few good ways to skin the approximate virtual cat, tossed ideas around and had good time. The initial C# vs VFP comparison was just the nail for the soup.

(The nail soup is an old Serbian folk tale, about a hungry soldier who was traveling home, and asked an old woman if she had some food for him. When she said no, he asked for some water and a pot to cook a nail soup. The woman got curious, and allowed him to cook it. He pulled a huge nail from his bag, let it cook for a while, then asked for some onion, parsnip, potatos etc... in the end he took the nail out and put it back into his bag. The old woman treasured the recipe.)

>Many interesting points, re relative performance of various techniques and other things, have been made on this thread.

[snip]

>I don't personally see the point of the cross-language comparisons being done here. IMHO they're invariably subjective and there are always tons of factors impacting what is the best tool for any given job, of which runtime performance is only one. But I'm not trying to inhibit you from those comparisons if they serve a purpose for you -- just indicating that I don't have anything to contribute on that subject ;-).
>
>>L<
>
>
>PS I realize that building XML documents does not have to be the goal of your tests. They were supposed to evaluate and compare speed of string handling, file writing, and iterating through data. But still, if your test case is supposed to create XML, please do a proper job.