From: "Tony Ingenoso" 

A loader would/should be smart enough to load what's needed.  Of course you
get whatever is on a page if you want it or not -- BUT that is where
careful design comes in.  Link so the most common stuff winds up on related
pages and keep the oddball cases, (lengthy) error paths and related data on
pages of their own.  If the loader is too stupid to load them anyway, a
pager should idle them out quickly enough that they have minimal long term
impact.

People are lazy though - few will ever look at the link map to see where
things are winding up.

I'd be willing to bet that the vast majority of commercial apps could have
their working sets cut almost in half with only a couple of days work
pouring over link maps and shuffling things around.

Look at how relatively well Warp3 ran in 8M.  That was a system that was
page tuned rather well by people who knew how to do it.

"Geo"  wrote in message
news:427d472e$1{at}w3.nls.net...
> Actually, it does mean larger because you can't have all tiny pieces. In
> other words a dll has to have a bunch of related functions and you have to
> include the whole dll even if you just want one of those functions. There
is
> a bunch of bloat in that kind of method but I don't see any easy way
around
> it where you could have reusable chunks of code, if you try to avoid that
> the result is basically the same thing as a high level language.
>
> Although.. I suppose a compiler could be smart and simply remove the
unused
> functions.
>
> Geo.

--- BBBS/NT v4.01 Flag-5