From: Philo, John <jphilo@amgen.com>
  To  : 'Borries Demeler' <demeler@bioc02.uthscsa.edu>
  Date: Thu, 10 Apr 1997 07:45:45 -0700

RE: binary files: the smaller size GOTCHA!

No, on DOS/Win95 hard drives you have no choice about the MINIMUM
allocation unit--it is determined by the fact that the FAT is only 16
bits (yes still).

I don't know anything about Mac hard drives.

I agree that the most legitimate reason for binary files would be if it
affects what experiments one can do.  However, I am pretty sure Walt is
absolutely wrong that writing the ASCII file slows things down.  Clearly
your examples show that with a good compiler this is not significant at
all.  I think Walt (and Tom on occasion) are confusing scan (fringe)
ASCII files and IMAGE files in ASCII (which is all the XL-I will
generate, and it is very hard to get them at all).

What do you mean by "correct the time" for the acceleration period.
This sounds dangerous to me.  I see no reason for such corrections.  I
know, for example, that Alan Minton's velocity software treats the
initial time as a fitting parameter, and tries to justify that with
words about acceleration, but the truth is that is just an empirical
fudge factor to make things fit better and has no basis in theory.

As to your data structure for velocity, I agree it makes sense, but I am
not clear whether you are proposing this as a standard, and I still have
grave doubts about binary files as the primary data format:

I have received e-mail from quite a number of people who say they are
not hard-core programmers, but are contemplating or beginning to do some
programming.  They are VERY, VERY concerned about binary formats and the
cross-platform compatibility problems they raise.  Yes, for you and I
and Walt we can deal with it even if we have to swap byte order or
whatever, but that is the kind of thing that could stop those folks
COLD.

Good idea to remind folks of your RASMB archive.

'Best,

John

>