What happened to WinFS?
I don't think anyone really knows.
The most up to date information I can find is an interview with Quentin Clark in which he says that "most of WinFS either already has shipped, or will ship" just in other forms, such as part of SQL server or the ADO.net entity framework.
I personally suspect it was one of those projects that was poorly defined from the start. It was all things to all people and consequently could never fully realise all it's goals. It seems that all the research that went into the various different aspects of WinFS eventually grew into separate projects and became parts of other things.
You can see from this development timeline that it's been an ongoing project since 1990. that makes it one of the few software projects that actually been in development longer than Duke Nukem Forever
[Edit: For completeness, here is some other info I found - WinFS Blog - Last updated June 2006. The last entry basically says WinFS was not dead but is no longer a separate product, it was planned to incorporate the tech into other products like SQL server and ado.net]
New information (May 2010):
I found this article which talks about the features of WinFS that have survived and live on in some form in Windows7.
It became part of SQL server 2008.
I worked on a team that had a partnership with microsoft to demonstrate the feature by integrating it with our product's data storage. there's a white paper floating about on the MS site (and frequently appears on the register) outlining the Marketing Bulls... uhh.. technology integration involved. if you look at my profile, that might give you a clue what to look for...
edit: I believe this is the feature it became
Just for the record, an excellent ZDNet overview from 2013 (titled "Bill Gates' biggest Microsoft product regret: WinFS"), with lots of relevant links, references and insider quotes like:
... no two members of the WinFS team seemed to be able to answer the question "What is it?" in a succinct, cohesive way. He noted:
"Some people would say that it (WinFS) was the database embedded in the operating system. Others would say that it was XML in the database. Other answers included a SQL-based file system, an object/relational mapping layer, rich storage for Office apps, a file system metadata indexer, .NET in the database, etc. It's not that these weren't useful technologies (many of them shipped outside the WinFS project), it's that there wasn't a singular vision driving the project to completion. Because no one could identify the essence of the project, it made it really hard to make any decisions about what should or shouldn't be included."
In his referenced blog post Welcker continued:
... it provides a barometer for other projects that may be headed towards certain doom. If you can’t boil down what your project is about (and everyone is be able to say what it is) then it’s probably not going to be successful. I would argue that even a single, incorrect goal is better than a wide variety of semi-correct goals. You can always change the course of a team headed in a unified direction but trying to corral a team headed in every direction is pretty much impossible.