Question: Solve using side relations

Consider the following lines:

eqs := {x - a*y,y - a*x};
sol := solve(eqs,y);

In general, sol will be NULL, but if a²=1 then the equations can be solved. How can such side relations be implemented in connection with solving?

PS: The above example is of course a grossly simplified one.

Note added: Perhaps it would be prudent to mention that the variable 'a' above is intended to be a symbolic place holder for some operator acting on some functions x,y, and having square equal to the identity map; it is not simply some algebraic number. The equations should thus be solved for x,y using a²=1, without having assigned anything to 'a' itself. For the simple example above, this can be done using the function eliminate, as suggested by Axel Vogt, but for more complicated cases, its success seems quite unlikely: for one thing, how can one determine, by looking at the set of equations, which variables can or should be eliminated?