OK so spot meter.
I've got a D200 so the controls might be a bit different.

Yeah, I've just begun playing with the spot meter. With practice, it
should be possible to get the desired results without checking the
histogram and I think that is the value. I am accustomed to using EC+/-
with 'normal' matrix metering but that's actually not so predictable. So
now I'm using the spot meter, AE-lock button (set a bit different than
default) and I dial in EC+2 or so on white areas to get that little bit
of blinking highlights. Of course there are times you want to blow the
highlights even more because the shadow area is important or to remove
detail in the highlights. In that case I'll meter with EC+0 on something
more neutral in the shadows. The other way to go is just put it in
manual and look at the meter reading in the viewfinder. That's handy
because you don't have to reset it for every frame if doing multiple
shots in the same lighting and no tricky fingerwork with the EC button.
But I never shot with a fully manual camera so it'd take a long time to
get familiar with shutter & f/stop options to where I really have any
clue where to start so I prefer aperture priority mode with EC.

The book you want is "The Camera" by Ansel Adams.

Basically what you are trying to do is to squeeze the huge dynamic range of
light that the eye can process into the limited dynamic range of the camera
and make the tonal relationships appear somewhat as they do to your eye. It
is an impossible task, so you have to choose the best compromise. As Paul
says, the first rule is not to "blow out" highlights by overexposing that
the eye can deal with. You don't want a face or generally even a white shirt
to lack tonal gradation, as this appears "wrong" since the eye can see tonal
gradation in a white shirt, so the trick is to expose so that just the very
brightest parts of the shirt at at 100%--pure white. OTOH if you have a
light bulb in the shot your eye will accept that being blown out, as
normally you don't look directly into a light bulb, or see tonal gradation
there--same with the sun.

But think about shooting a scene in a city where part of a light-colored
building is in full sun with a street in deep shade. If you expose for the
light-colored building, so that you have detail in that your street turns
unnaturally dark (according to what your eye can see). If you open up so
that the shadow detail in the street is a bit better you lose detail in the
building--it turns into a white monolith with just a bit of the darker
detail showing. What to do?

You will find in this situation that it is desirable to keep the highlight
detail, even if it means losing detail in the shadows. Both are ugly, but
the white monolith is uglier, and seems more unnatural.

There are various ways of dealing with the limited dynamic range of photos.
Film people use the Zone system, with selective exposure and development
variation to increase or decrease contrast, or masks post-facto to do the
same thing.

With digital there are various ways to expand detail in the shadows, but
there is really no way to get lost highlight detail back. It is virtually
always better to lose detail in the darkest shadows than to lose detail in
the brightest non-specular highlights.

Your example of metering a backlit sunset is one of those cases where you
are going to lose one way or the other. Even though your eye can see both
rich color in the sky and detail in the ground you won't get both on film or
in a digital file. Most people choose to expose to have the rich colors of
the sky and just let the foreground turn black, as a pale, washed out sunset
sky is not going to win any awards. Another possibility is to take two
identical shots (tripod!)--one exposed for the sky and one for the
foreground and combine them digitally. You can also use a graduated neutral
density filter to darken the sky in relation to the foreground, but there is
no way that without some tricks you are going to get it all in one shot the
way the eye sees it.

Digital cameras such as the D50 have a number of built in algorithms to
produce the best possible compromise exposure for various situations such as
the ones described above, and generally matrix metering works pretty well,
but it isn't perfect. But the great advantage of digital is that you have an
immediate preview and can (hopefully) reshoot the scene after you get some
visual feedback from the LCD monitor. This eliminates much of the need for
fancy tricks with film (such as spot-metering all the tonal values in the
scene and calculating exposure from that information), where you only know
if the exposure is decent long after the possibility to reshoot is past.

Toby