I have had the same thought numerous times now. I still think it's a great idea, but I also think that there is a very good reason that the current selection of online brew recipe software is as limited as it is. Quality web software that looks and feels like it's a native app is far more challenging and time consuming to write than a standalone client app. My full time job is a web developer and I honestly think that replicating all the advanced features of an app like BeerSmith on the web would take a team of 2-3 developers working part time the better part of a year to complete.
Also, I wouldn't even dream of starting such a project until the BeerXML 2.0 standard is complete. http://www.beerxml.com/
Once that is complete all the schema work is done. We would just have to create the same schema in SQL, MySQL, or whatever. Plus, if there is a real modern standard then that would give users the ability to upload their existing recipes into the online database, or download recipes and import them into client apps like BeerSmith.
This is personal preference, but I would only undertake a project of this magnitude using an object oriented language such as C# or Java. Yes I know PHP5 supports classes, but it still isn't a true OO language.
That being said, I wish you the best of luck. I am very curious as to what you will create. If you want a second set eyes or some help feel free to PM me.