Light has no rest mass. But, light can never be at rest.
Light does however, have momentum. And when it is reflected/refracted, it will transfer some of that momentum to whatever it hit.
Also remember that forces act on pairs of things. Light can be deflected by gravity (which will change it's momentum); therefore it must produce a gravitational field (which will in return change the momentum of whatever deflected it). And of course gravity is proportional to mass.
Light which is moving towards and object will gain momentum, like a falling rock would. Only instead of moving faster (since it's at a fixed speed), it shifts towards higher frequencies. If you look at the equations slightly differently, you get gravitational time dilation.
If you take a large-scale view, you get really good results by using a mathematical model that says that the light is behaves like a wave with a certain speed, and that the speed is slower in water.
If you take a small-scale view, you get decent results by using a model that says that individual photons have a small chance of being absorbed by the water molecules and then new photons are radiated again.
Supposedly there is a quantum mechanical view which applies on all scales and gives even more accurate values, but I have never seen it worked out for something as complicated as the interaction between light and a surface of water molecules. Both of the others I have worked out myself back when I was a physics student.
In a convex lens, photons take longer to pass through the thickest part, which creates a concave wavefront that naturally converges on a focal point:
http://arachnoid.com/example/index.html#Lens_Example
Photons don't need to have mass for this to happen, because they aren't being deflected like billiard balls, they're being slowed by their interaction with atoms.