Turing machine with k-tape, tape of output

Question

Consider a Turing machine with input alphabet $\{a,b\}$ that computes the following function:

$$ f(w, v) = \begin{cases} w & \text{if } \operatorname{length}(w) > \operatorname{length}(v), \\ v & \text{otherwise}. \end{cases} $$

I wish to use a TM with two tapes, the first of which contains the input string, say encoded as $*w*v*$. I'm not sure, having read several books, which tape is the output tape in this case. For example, the output tape for a TM of two tapes is the input tape or the second tape? If I use three tapes, is it correct that one of them contains the input, another is a working tape, and the third is the output tape?

Answer 1

There is no generally accepted definition of Turing machine. Various authors use various models. One author's definition might specify the output tape, another's might leave the choice to the Turing machine designer.

The reason that we don't care about this "calling convention" is that it doesn't matter from the point of view of computability (and, to a large extent, from the point of view of complexity). All of the different variants of Turing machine are Turing-equivalent; the languages they decide are the computable languages, and the languages they recognize are the recursively enumerable languages.

If you're writing a paper, the choice of which model to use is up to you. But this hardly ever comes up, since Turing machines are usually described very informally. The informal description can be implemented in any reasonable variant of the Turing machine model.