I found myself reading Les Orchard's
HexOddities wiki page in the
"0xDECAFBAD" world. There's a pointer to
a copy of a comp.lang.java thread where Bill Venners
asks about the magic identification number in Java .class files.
I sent Les and Bill a message with roughly the following couple
of paragraphs adding my little bit to that history-mystery:
I worked for Sun from about 1986 until 1995, mostly on debuggers
and then C++ compilers. Before Sun purchased Tau Metric, whose code was the
basis for the C++ compilers that they have sold since about 1991, we had a
small team start to implement a C++ compiler front end (intending to hook it
into Sun's common Intermediate Representation). I was the one who came up
with the name for that project: "CAFE" as an acronym for "C++, A Front End".
(That would have been around 1989, I think.)
We only worked on this for a few months before buying Tau Metric, a company
who had an already much more complete C++ compiler.
When we bought Tau Metric, the project name "Cafe" was kept,
but now referred to that compiler project.
Once that compiler was a Sun product, it had to have a marketing person.
That person was Kim Polese.
I can easily see people describing her as a "babe",
and she was the "Cafe" product manager.
She later transferred to FirstPerson, the semi-secret Sun
subsidiary that developed Oak, that turned into Java.
I believe that Naughton's answer isn't really a complete answer
about where the word CafeBabe came from.
At any rate, it's Kim that comes to my mind when I think of that term.
Then Dave Winer pointed to a picture of Kim Polese, next to
a mention of the 0xCAFEBABE magic number, and to
Steve Zellers' readings from the Unix /etc/magic file.
That file lists Mike DeMoney as the guy who chose "cafebabe".
(Twice. For two different kinds of file.)
I suppose this could be a complete coincidence; I guess
I ought to try to track down Mike DeMoney and see what
he might admit to.
Java history buffs might also be interested in
"Long Strange Trip to Java" narrative.