Question: I'm curious: How does convert/rational work?

I recently stumbled across this video Why do calculators get this wrong

At some point, the author (Matt Parker) mentions the Farey's algorithm as a way to get a rational approximation of a number between 0 and 1.
Just for fun I applied this algorithm to build increasingly accurate rational approximations of Pi (see here Rational_approximation.mw). 
The rational approximation 

convert(evalf(Pi), rational);
                             104348
                             ------
                             33215 

is one of those the Farley's algorithm returns.

Is this pure chance, or does Maple indeed use the Farley's algorithm?
If not, what algorithm does convert/rational use?

Please Wait...