> > 	the reason for picking 32bit timeval is, I think, as follows:
> > 	- there are too many existing files with 32bit timeval, and
> > 	- there's no robust way to determine 32bit timeval from 64bit timeval
> > 	  so we can't really read in both
>Uh, I don't understand.  Why can't they just add some backword compat
>mechanism, change the magic number on the file, etc?

	I'm not 100% sure either.  it was not me who did this part...

itojun, one of tcpdump-workers@tcpdump.org