Take the spring equinox (spring in the Northern Hemisphere, of course); it 
is usually on the 21st of March.   I used to think that this was the day
when day and night were equal [from the derivation of 'equi' and 'nox'
(for night)].  But I now know that on the 21st of March, as given by
sun-rise and sun-set tables, the day is something like 25 minutes
longer than the night.  [London sunrise 21 March 2005 was 6:00 a.m.
but sunset was 6:13 p.m.]

I've been offered two explanations for this.
1.  that sunrise is the leading edge of the sun coming up over the
horizen and sunset is the trailing edge of the sun dropping below the
horizon, and that if you measured the time from the diameter of the
sun at each end of the day the time would be close to equal.
2.  that the light of the sun is bent by going through the atmosphere
at sunrise and sunset and so one is actually seeing the sun before it
actually would be above the horizon by measure of a straight line.

How much do each of these factors effect the timing of sunrise and
sunset at the equinox?  I've seen some explanations of this (NY Times
Science column a few years ago) where #2 is the only factor mentioned,
and others where only factor #1 is mentioned.

Is one the predominant factor and the other only a second-order effect?

---

Clarification of Question by gregorydove-ga on 31 Jul 2006 11:40 PDT

I the two answers given you exactly pinpoint my question but not the
answer.  First of all, the discrepency is not 'a few minutes':  the
day is 26 minutes longer than the night! And my question remains,
which of these two factors accounts for the lion's share of the
discrepancy.  My 'gut-feel' is that the disc vs. point of light would
account for 90 to 95% of the discrepancy, but I have someone betting
with me that each factor [disc. vs. point and bending of the light]
account for about 50% each.  So I'm looking for the definitive answer.

Normally the effects of the atmosphere have more of an effect than the
leading edge of the sun coming up before its centre.  The only
exception would be close to the poles.

This US military FAQ shows that normally atmospheric refraction
effects sunrise /sunset times twice as much as the sun's edges
appearing before its center.  This would be considered "definitive".
http://aa.usno.navy.mil/faq/docs/equinoxes.html
"The times of sunrise and sunset in almanacs are calculated for the
normal atmospheric refraction of 34 minutes of arc and a semidiameter
of 16 minutes of arc for the disk."

Keep in mind that ?minute of arc? is referring to 1/60 of a degree
rather than an actual minute.

Here is some further explanation of my answer:

This article states that the upper limb of the sun coming up only
account for about 2 minutes or your discrepancy. (at the equator) 
This would only make it a difference of 4 minutes between "day" and
"night".
http://en.wikipedia.org/wiki/Equinox
"Sunrise and sunset are commonly defined for the upper limb of the
solar disk, and not for its centre. The limb is already up for at
least one minute before the centre appears, and likewise sets one
minute before the last appearance of the limb sets too."
This effect becomes greater as you move away from the equator.

Your latitude determines how much of an effect this will have.  London
has a latitude of about 51 degrees North.  It will see more sun than
someone at the equator because of the angle at which the sun sets.
(see the cool pictures on the wikipedia page)  This increases the time
you see the sun because it sets "slower" due to the angle.  This of
course gives more weight to your side of the argument, but not enough.
 The only time you would be right would be when you are much closer to
the poles.
http://en.wikipedia.org/wiki/Equinox
March 20, 2005 at the equator = 6:04am to 6:11pm

London is at about 51 degrees North, so the ratio you are looking for
will be very close of 50/50 if not leaning slightly in favor of
atmospheric refraction as the leading cause of this phenomenon.

As a point of clarification:
The equinoxes varies between from March 20th and 21st and September
22nd and 23rd.  In 2005 the Spring Equinox was March 20th.  The
sunrise / sunset times were: 6:03 AM and 6:14 PM.
http://www.timeanddate.com/worldclock/astronomy.html?n=136&obj=sun&month=3&year=2005&day=1

If you require any further information please post a request for
clarification and I will respond as soon as possible.

Thanks!

Well done, once the clarification was made.  Thanks.

Someone else was quite satisfied with the answer to a very similar question:
http://answers.google.com/answers/threadview?id=495777