S-units and compact representations in number fields
Huge algebraic numbers are problematic because- computing with them algebraically is expensive;- approximations via floating point embeddings into C require huge accuracy (cancellation);- they are often intermediate results: we do not want a result in K but in K∗/(K∗)2, or in- ZK/pk, or a floating point approximation to complex embeddings, or . . .- they may overflow the possibilities of the implementation: try 22100 .