Hi Duncan, > Computer scientists (which I am not) will possibly recognise this as > a Lin-Kernighan type algorithm. IIRC, a copy of the original paper by them is available from Kernighan's home page, perhaps that has some implementation details? Ralph.