What's magic about these numbers?

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

I have a savefile I want to upload to my IFS via TCP/IP somehow.

The savefile is 4,294,966,104 bytes long.  This in itself is an odd number,
being only 1192 bytes LESS than a "true" gigabyte, 1024^3, or 4,294,967,296
bytes.  But hey, the size of the file is the size of the file.

In any event, no matter what I try, the most I can seem to get up to the IFS
is 4,293,918,720 bytes, which is exactly one meg (1,048,576 bytes) shy of
one true gigabyte (I will refer to this number as "1GB-1MB").

1. Using a normal FTP to a savefile, I get an invalid record right at the
end (after a long transfer), right around the 1GB-1MB point.  I'm not
exactly sure of this number.

2. Using FTP to the IFS, it thinks it completes successfully, but the target
file has only the "1GB-1MB" number of bytes in it.  This number I'm sure of.

3. The jar utility can't handle files that large.  It think the file
is -1192 (that is, negative 1192) bytes long.

4. Using Windows copy or DOS copy, the connection drops anywhere between 1GB
into the file and almost the end, usually before 2GB.

5. I copied the file up to the IFS in four chunks, each of which were
1,073,741,526 bytes long (exactly one fourth of the original).  When I
copied them together using "cat" under QShell, the resulting file was: you
guessed it!  1GB-1MB long.  Again, this number is certain.

Checking, I found an IBM document
(http://www.as400.ibm.com/tstudio/tech_ref/syslimit/limitv45.pdf) purporting
that the supposed limit for IFS files in V4R5 is 256GB.  But I also know
that 4GB is a magic number, being exactly 2^32.  My question is this: is
there really a "phantom" upper limit of just under 4GB to file sizes on the
IFS?

Joe