The name of the site is something like OnStreet.com. (Which is currently squatted.) Its purpose is simple: show me the nearest legal parking to where I want to go. The user interface is probably a Google Maps mashup. I select a location on Google Maps; tell OnStreet that that's where I want to be; and it shows the parking options as a map overlay.
What makes this different from most of what I've seen so far is that the data is crowdsourced -- you encourage people to add info to the map about what is legal and what isn't on the street, and provide easy tools to, say, mark that this location that claims to be legal no longer is. Not clear exactly how you get people to add data, but at the least you provide public credit and ego-boo for the people who add lots of info to the database. Of course, it should also let you describe parking lots and garages, with info about how much they cost, but cheaper on-street parking is the heart of the idea. (Indeed, garages might need to pay a little to get into the listings.)
It should grok the notion that certain spaces are only legal at certain times of day, and have a way to say that, say, this block has three legal spaces, as well as how much the meters cost.
You should be able to easily specify filters describing what you are willing to cope with when looking for parking -- whether you prefer to pay more money, or walk further. (And how far is "too far" for you.)
The thing should have a mobile app if at all possible, because that's where it would really sing. For example, it should have an Android app that adds a button to your main screen; push that, and it means "show me the nearest legal parking to where I am right now", and pulls that right up as a Layer in Google Maps -- instant, super-useful gratification. If possible, it should hook into Google Maps Navigation, and actually direct you to the nearest places that might be legal. There should be a big button that says, "No, there's nothing here -- take me to some other alternative". It should keep track of that information and build a statistical database -- if there never seems to be available space here at a particular time of day, stop routing people there. Ideally, it should keep track of when it last sent someone to a particular location, and have a sense that that location is now more crowded.
It should *not* keep any user information. Frankly, it doesn't have to: there's plenty of money to be made just from the statistics, and the immediate knowledge of where the user is. So in these days where people are starting to actually get a little privacy-conscious, you can make a big deal about it *not* doing any sort of creepy tracking.
The best part is that it has a rock-solid business plan. It's advertising supported, but can charge a *hefty* premium for advertising, because it's strictly location-based -- it shows ads for businesses that are close to the specified parking space and target destination, and between them. Businesses buying advertisements can specify filters for things like how far is relevant for them, and what hours are relevant, so that the service is less likely to show an ad for your service if you're closed. Ideally, the ads shown are slightly randomized (or at least, rotated), taking time and location into account. The plan might have an auction or sliding scale, so that you pay more for a parking space right in front of your business than for one two blocks away.
For extra points, it might have extremely simple search capabilities built in. For example, a button for Food would highlight all the food-related advertisers near here. You could charge quite a lot extra for that.
Okay, yes, there's one problem with the business plan: Google aren't stupid, and if this seems to be making money, they'll eventually build it into Google Maps proper. But even that is potentially fine. The service is theoretically decently cheap to build and run (there's some serious programming involved, but it's probably much easier than most of the projects I do). So you set it up, make some modest money for a while, and your eventual likely exit strategy is to get bought by Google for at least some good kicking-around money.
That's the heart of the idea. I suspect it's not too terrible to implement, and would be popular in these cash-constrained times -- giving people a good tool for finding onstreet parking (rather than just expensive garages) seems to fit the zeitgeist. I'm rather too employed to do much with it right now, but if one of my friends wants to run with it, go for it -- just give me some credit for the idea, and maybe some options if it goes anywhere...