MEMORY

[ANDREW GLASS]

Robert:

I concur wholeheartedly with your comments. Sure, I'll give it an open-minded
look. It wasn't so long ago that I thought a Tandy 8K color computer and my TV
set were as the cat's whiskers.

With the new generation of larger programs, memory becomes a vital
consideration. I can live with any oversized executable file on a 322MB C:>
drive. But DOS memory is a precious and easily wasted commodity.

I've become quite spoiled employing XyWrite as a "shell." With major help from
David Auerbach, I'm able merely to type ss at the command line and juggle
salaries in Quattro 2.0. Or I type db and crunch a Paradox 3.5 campaign
finances database. (Less frequent needs are easily met, essentially in the same
manner, via your oodles library [Tab-Enter], which sits in RAM-like splendor in
D:>.) These painless exits and re-entries (without any mediation from Desqview,
Carousel, Windows and that breed) are possible because of the way Borland has
implemented its virtual memory manager in these releases. Another essential
element is 386^Max and a 4MB platform.

The big unknown is how XyWrite IV will implement its own memory management. We
need to know whether it will make good use of expanded memory. We need to know
if it is modular. They don't want me as a beta tester. So I will have to wait.

As matters stand, it is imperative to use a load/unload "sandwich" when
invoking any TSR, no matter how small or well behaved, in the stack above
XyWrite. Otherwise, XyWrite's overlays will usually become confused and drop
into a black hole. I do this now with SuperKey, a notoriously difficult TSR,
using a utility called FMARK / UNMARK. That does the trick.

By the way, on the road I use a ROM version of XyWrite with my NEC Ultralite as
B:> (It should survive the Saudi winter if Saddam does not bomb it. Then the
president can take me to Indonesia in February.)