Remote Method Invocation and Object Serialization - Java: The Good Parts - page 160

[ 31 ] Well, it is now. In the early days of RPC systems, how to determine the byte order of primitives was
a subject of considerable debate. So there has been some progress in computer science.
[ 32 ] What to call the process of converting an object into a stream of bytes that can later (and perhaps
elsewhere) be used to reconstruct a copy of the object was, at one time, a subject of some dispute. One
of the common ways to refer to the process was by using the terms pickling and unpickling . Prototype
versions of the Java process followed this naming convention, but this was objected to by some mem-
bers of the Java team (the passive voice is used to protect all involved). After considerable discussion,
the terms object serialization and deserialization were agreed upon.
[ 33 ] I have tried to write this without using the pronoun we a number of times, and it just doesn't work.
Object serialization was first developed by Roger Riggs and Krishna Bharat while they were in Sun
Labs, in the same group as Ann Wollrath (who at the time was developing RMI). I was the technical lead
of the group, and we all discussed the work that was going on. So in this case, when I say we, those are
the people who I mean, along with Ken Arnold, also part of the group, who was writing a book at the
time.
[ 34 ] See Andrew Birrell et al., Network Objects , http://birrell.org/andrew/papers/115-NetworkObjects-
SOSP.pdf . Note that two of the authors of the Network Objects system were also the authors of the first
RPC system, on which much of the original CORBA RPC system was based.
[ 35 ] Actually, the assumption in RMI is that both the sender and the receiver are written in such a way
that they have a binary made up of compliant Java bytecodes. At the time that RMI was first designed,
this was the same as the assumption that both the sender and the receiver were written in Java, but as
more and more languages compile to Java bytecodes, this is less and less true.
[ 36 ] Those interested in these problems can do no better than to read the Sun Labs technical report
by Michael Warres, Class Loading Issues in Java RMI and Jini Networking Technology , ht-

tp://research.sun.com/techrep/2006/abstract-149.html .