Skip lists, are they really performing as good as Pugh paper claim?

Question

I\'m trying to implement a skip list that perform as good as a BST using a minimum additional memory overhead, at the moment even not considering any memory restriction the perf

Answer 1

I doubt that the skip list was the better choice than an e.g an AVL tree even in 1989. In 1989 or 1990 as a student I implemented both: it was not a good implementation of the skip list, I must admit, I was a novice in that time.

However the AVL tree was not anymore difficult to implement. In contrast I had difficulties with that variable length forward pointers of the skip in list implementing in modula 2, which I primitivley solved with always using a maximum of 16 next pointers.

The advantage of fewer operations in inserting, I never saw. The AVL tree, if I remeber correctly had on average of about not more then 2-3 operarations. So the expensive rebalancing happens not often.

I think it was more a hype.