How many extra (leap) seconds have we found it needful to put in over the years?

The web page provided by the U.S. Naval Observatory ("The Official
Source of Time Used in the United States")
http://tycho.usno.navy.mil/leapsec.html discusses this in detail.  It
references a list of leap seconds, and the dates that they were added.
I counted 21.

The program of leap seconds only began in 1972, to insure that the UTC
(Coordinated Universal Time.) [1] differs from UT1 by less than .9
second.  UTC is based upon an atomic clock, whereas UT1 is based upon
the Earth's rotation.  UT1 is not uniform.

At the time that the leap second program was started, we used a
standard length of day from 1900, so the day was already about 2 msec
longer by 1972.  So, after about 500 days (I use this value just to make
the arithmetic work out), a leap second would be added.  If the Earth
never slowed at all again, we would still have to continue to add leap
seconds.  It would probably be a long time before we'd have to subtract
one.

Actually, the Earth does not prevailing slow down.  Seasonal
fluctuations are strong (changes in moment of inertia due to weather
patterns), as well as tidal fluctuations (monthly changes in moment of
inertia due to the uneven distribution of continents), but those
compensate for themselves over the course of a year or month.

Earth rotation is often described as steadily slowing, but it is not,
even over thousands of years.  Historical analysis using timings of
Babylonian eclipses from 2500 years ago give different values than
Arabian times from 1000 years ago.

There are other geophysical analyses, but hey this is rec.puzzles.  If
you want to talk about it more, let's go to alt.fan.cecil-adams (who
doesn't seem to have discussed the matter [2], I'm shocked.)

[1]   So why is it called UTC and not CUT?  That's a very funny story:
      http://www.bldrdoc.gov/timefreq/faq/q5.htm
[2]   http://www.straightdope.com/topics/index.html

-- RM Mentock <mentock@mindspring.com>, May 1998