Nested loops is one way of processing joins:
for each row of table A
if this row matches where clauses
for each row of joined table B
if this row matches where clauses and join clauses
accept row
end
end
end
end
That can be optimized with indexes quite a bit, by doing "for each row found at key K in some index" instead of "each row of table A", and the same with table B.
The presentation is saying this is the only way MySQL processes joins. There are other methods than can be used, but MySQL doesn't implement them. This OraFAQ entry gives several that Oracle implements: http://www.orafaq.com/tuningguide/join%20methods.html Similarly: http://oracle-online-help.blogspot.com/2007/03/nested-loops-hash-join-and-sort-merge.html
"Everything is a join" is just an implementation detail, I believe. Not really that important.