Removing Redundant Rows?

Hi All

Can anyone help me with this problem?

I have an 11 million row table that lists relationships between pairs of people. There are several columns, but the relevant ones for my problem are

personA, personB, a-b_strength, b-a_strength.

because of the size of the table i'm having all sorts of problems, and i could actually cut out half of the rows, because for every row listing eg

dave, steve, 4, 5

there's another row listing the same information, but flipped round, ie

steve, dave, 5, 4

I'm sure there must be an efficient way to delete, using a query or otherwise, one of each of these pairs (it doesn't matter which). I have another column (id) that simply has a unique number for every row, ie 1,2,3->11 million, so i was thinking of something like

delete
from table t1
where exists
(select *
from table t2
where t1.personA = t2.personB
and t1.personB = t2.personA
and t1.id < t2.id)

So, this should delete every row that has a mirror version earlier in the list.(i think)

What i'm worried about though, is performance: is this going to make a new temporary table for the nested select for each new row in the outer statement? I already created a table that listed the count of friends for each unique person with a make table query, and that took 12 hours to complete. I don't mind another 12 hour query, but not a 144 hour query!

Any advice welcome - thanks! (Sorry for the longwinded post)