The approximate solution of the m-machine problem is addressed. The La
m-Sethi worst-case analysis of the Coffman-Graham algorithm is set up
to be partly incorrect. A slightly different context is set up to corr
ect and complete this analysis. It is shown that the makespan of a sch
edule computed by an extended Coffman-Graham algorithm is lower than o
r at worst equal to (2 - 2/m)omega(opt) - (m - 3)/m, where omega(opt)
is the minimal makespan of a schedule.