re: [SLE] tar performance varies wildly