dn> Shift your ctime value left by 3 bytes

 dn> Err, _bits_, not bytes, of course. :-)

Yeah, it took a while for that to make sense!  I was going to wait until
I'd done a calculation, but I expect >8 file opens in a second is not
unreasonable.  I want a few more bits shifted (I like your idea!), and want
to see how many bits I can screw out of it.  I expect I need the time value
to cater for 48*60*60 << x and still be less than FFFFFFFF.  So let's
see FFFFFFFF = 4 billion.  48*60*60 = 172800.  So 4 billion / 172800 =
24855.  24855 is the next number above 2**14, so I can afford to shift 14
bits, or 16384 files per second.  Since I think 1024 files should be
enough, I'll shift 10 bits instead, and gain unique files for 16*2 = 32
days.  What do you think of that?  BFN.

Paul

---