“…One procedure is Sylvester's theory of eliminants [19], but again we quickly encounter polynomials of very high degree [11], and the method is ill-conditioned for even moderate p, rn and n. An alternative is to define P 2 n-1 (R) by P()=-=1 p()2, and to set f= ('zpP)]S"-1. Then f zP(R)IS"-I. Construct an evaluation tree for 504 GEORGE E. BACKUS (Sn-l, , f, qw, a, r) as in Theorem 2.5, with qw as in (5.5) (and p replaced by 2p) and with limi_,oo a (j) limj_,oo r(j) 0.…”