In his book Galois Theory, Ian Stewart mentions:
In $1813$ Paolo Ruffini published yet another version of his impossibility proof. The paper appeared in an obscure journal, with several gaps in the proof (Bourbaki Elements d'Histoire des Mathematiques $1969$, page $103$). The most significant omission was to assume that all radicals involved must be based on rational functions of the roots.
I do not understand the last statement of above quoted paragraph - radicals involved must be based on rational functions of the roots.
Can one explain with example, what Ruffini specially assumed in his proof, which is not correct for the proof of his assertion (of insolvability by radicals)?