The problem with using chemicals in water for digital information is that the chemicals would need to arrive at the tap in the same order they were sent. While this will sort of occur during conventional fluid dynamics, the loss of precision would mean that you'd not be working with a very high bandwidth on this at all.
That said, there is a solution; Laminar Flow. In fluid dynamics, laminar flow is the liquid equivalent of lasers. It affords coherent flow of fluids through things like water pipes and the like, and as I understand it many of the fire fighting nozzles for hoses are designed to generate laminar flow to make the water go further as a coherent stream from the nozzle than would normally be possible before it turns into a spray or mist.
You could use chemicals in this flow, but if you want to preserve the ability to drink the water at the same time, you could experiment with some form of pulsed laminar flow from the tap. Water glasses would fill but the water would be coming out in very fast pulses, which if you get the system working well may look similar to a conventional flow of water in any event. BUT, a sensitive pressure valve may be able to read the intermittent slight drops in pressure as 0s and 1s, allowing you to transmit data over the (tap) line.
To be sure, this would require a very sophisticated plumbing system to be in place, but the reality is that without laminar flow, your chances of putting the chemical sequence together at the other end would be practically non-existent. Additionally, pulsed laminar flow could be detected via a pressure valve on a constant basis, meaning that so long as you have water flowing from a single tap (designed for this purpose) you can record the data being transmitted. That water could then be stored for sending back to a central location, allowing for a user to transmit data back via a 'return' pipe. It would also mean that you should (in theory at least) still be able to use your plumbing system for its original intent as well, which is dispensing water for household use.
I want to stress that I know there are some BIG holes in this model. Drawing a bath may well alter the pressure on the valve (although it should be able to detect relative pressure pulses), water pressure on the return trip may not be very high (unless you pump it into a gravity feed tank, kind of like a water antenna) and unless you have a separate supply pipe for every home, you're going to have a massively difficult time doing the equivalent of IP routing. Still, if it's all you've got, these problems are really engineering matters that would get solved with sufficient application of ingenuity.
All in all, water transmission would be difficult, but not impossible.