-=> Bob Jones wrote to William McBrine <=-

 BJ> Since I expect to only run into two's complement machines now,

!

Wow, one's complement... I have to admit, that's something I never
considered. Now that you mention it, I think I can probably go on ignoring
it. ;-) But you do have me thinking about how I'd implement those functions
in a compatible way...

BTW, does Max often store negative numbers on disk? I would expect that to
come up seldom, if ever.

 BJ> integer data can easily be handled as you suggest, maybe even with a 
 BJ> union on the structures.....

A union kind of negates the purpose of avoiding the "packed" attribute. 
(Not that that was the only purpose, but still.)

IMHO there is really little point in retaining the old methods; it just
makes the code more complicated (with the big-endian side likely to end up
less well-maintained), for very very little performance gain. (If you want
to maintain dual versions of the structure definitions for the benefit of
third-party programmers -- as in bluewave.h -- that's another story; you
can do that easily. But I see no benefit to having dual access methods
within the Maximus code itself.)

... I had to hit him -- he was starting to make sense.
--- MultiMail/Linux v0.45