Quote:
Originally Posted by xilman
I
If the rumour is correct, then perhaps we can deduce that the algorithm requires finding O(sqrt(N) smooth integers of size bounded by O(sqrt(N)) . I've been thinking about such things for a little while now  unsucessfully it's needless to say.

What you describe sounds like a N^(1/4) complexity algorithm, which would be much slower than L(1/4) complexity.