In the original paper, ElGamal(1985) starts his discussion of signatures by saying:
"The signature ..... is chosen such that the equation:
$\alpha^m = y^r r^s \text{mod p}$ (equation 3 from the paper)
is satisfied."
(In this paper: $m$ is the message, $s$ is the signature, $y=\alpha^x$ mod $p$ is the public key, $x$ is the private key, $k$ is the ephermal (secret) key and $r=\alpha^k$ mod $p$)
Everything else follows from this. Why this particular form? What the rational if any? This just seems to come out of no where. Why not, for example:
$\alpha^s = y^r r^m \text{mod p} $
with $s = rx+km$? (There is probably an obvious reason that I am missing, please do point it out!)