It's tempting to riff on the topic:



At 06:30 PM 8/31/2006, Mike wrote:
No, the word "set" has about 43 meanings, and none of them is wrong. You
just have to be clear about your context. Unfortunately, google isn't.

Admittedly it's confusing, but one usage of "top-level element" is a 
non-technical usage within the domain of XML (wherein it designates 
what we also call the "root element" or, for those who wish to avoid 
a terminology clash in XSLT, the "document element"), and another, a 
technical usage within the smaller domain of XSLT.



It won't be the first time there has been a semantic conflict between 
usages in languages or technical terminologies. Far from it: they're 
as old as language itself. Such conflicts are usually not resolved 
(and almost never permanently resolved) by dubbing one of the terms 
"wrong". Instead, we all learn to rely on contextual cues (as Mike 
points out) to disambiguate the possibilities.



Indeed, "meaning" (it can be argued) is never really inherent in 
terms or tokens, but is only incidentally instilled or applied to 
them, and always depends on an informing context. Such contexts have 
a way of changing; nor can they generally be frozen for long. "Dead" 
languages are those whose natural contexts are gone; technical 
terminologies are those whose context is insulated formally by a 
discipline or technology but which are "called in" to natural 
language to make their worlds open for our inspection.



This simple but inescapable fact about semantics in the real world is 
one of the most besetting problems, if not the fundamental problem, 
in the development of systems that seek to process information 
without doing the only thing that ever works to get things moving 
again whenever they get stuck -- namely, paying attention. 
Unfortunately, there seems to be a myth in our culture that meanings 
are inherent in signifiers, and that automating intelligence is 
merely a matter of chaining tokens together into the correct logic 
circuits. While in theory, given a sufficiently controlled processing 
model, much might be achieved this way, those who credit this myth 
absolutely are destined to spend their lives banging their heads 
against the wall.



Back to XSLT ... if the myth were true, we might have automated this 
list long ago.



Cheers,
Wendell