WinFS no longer a product

DCMonkey wrote:

Those that don't seem to think:
1) Spotlight did it first

It isn't that Spotlight did filesystem queries first (I thought it was BeOS*), but rather that they delivered at all without all the hype and leading developers down dead-end alleys.  Whoever it was that said that Microsoft should not say anything until it is certain that they can deliver hit the nail straight on the head.  Why do Apple fans get so excited about Apple stuff? Because they constantly surprise us by delivering cool things that haven't been hyped to death before hand ...


 * Interesting note, the author of BeFS (Dominic Giampaolo) now works for Apple. On the spotlight team. I wonder how many Be Engineers actually ended up at Microsoft. Anyone know?

The point I was kinda trying to make was :
1) Spotlight did it first (and Apple knows how to design software because they had a HIG 13 years ago, therefore anything MS might be working on (like WinFS) could only be a poor copy of it and couldn't possibly offer anything additional that might be worth investigating and implementing for Linux)

and

2) ???* (Beagle works like Spotlight and therefore is as good as Spotlight)

therefore

3) Linux is better than Windows (Beagle is better that WinFS).

And so no-one would bother to build a WinFS clone.

And yes, having Dominic onboard ( I think I mentioned him earlier in this thread), is one of the reasons I think Apple has a better chance of doing something approaching WinFS.

If they do, then maybe the Linux camp will have a better roadmap to follow (once they get past the ReiserFS vs SQL database, MySQL vs PostgreSQL vs pluggable back end, C vs C++ vs Mono vs Java, DIY schema vs Dublin Core/sematic web flame wars.

DCMonkey wrote:

If they do, then maybe the Linux camp will have a better roadmap to follow

They do like to follow ... come on Microsoft, how bout BeFS on Windows? It isn't perfect but it is closer than where we are now.

Rossj wrote:



DCMonkey wrote: If they do, then maybe the Linux camp will have a better roadmap to follow

They do like to follow ... come on Microsoft, how bout BeFS on Windows? It isn't perfect but it is closer than where we are now.

Given what is in Vista right now, the main thing I can think of from BeFS that I would like to see in Vista search is the ability to add one's own metadata to files that don't already have that metadata. Right now, you can only search on whatever metadata fields a file format already contains and that the IPropertyStore written for that file format by MS or a 3rd party decides to read/write. This means, for example, that you can't put your own metadata like "project name" with a value of  "FooPro 2.0" on a plain text source file in Vista.

Now if someone were adventurous, maybe one could write a stable of property handlers that could slip extra metadata into the system but store it in some database, but it would be nice to have it built in.

The upside of Vista's current implementation (and Spotlight's for that matter), is that you wouldn't have to worry about your metadata becoming separated from your files when you copy them to other systems, as you can always rebuild your indexes from the files. IIRC, with BeFS, you or your apps had to do any synchronizing of internal file metadata and file system metadata themselves instead of the system doing it for you by calling plugged-in handlers.

So while MS could certainly learn from BeFS, I don't see them switching to it as having any real advantage.

Some Microsoft ex-employee:

http://gdayworld.thepodcastnetwork.com/index.php?p=236

Interesting bit of history though, I was actually in Building 25 of the Exchange team, when Bill officially told Brian Valentine (he was running the exchange team at the time), that it was going to be this combined filesystem, he had a [inaudible] for about the third time while I was at Microsoft. 'cuz up until that point, the Exchange store was the file service du jour for unstructured ad-hoc data.

Very interesting problems, having trying to design a database that can handle structured high-transaction integrity data like a database, and to do the unstructred soup like a message store. It's still too frackin' hard; no one's figured out how to do it. And here we are, that's what really killed it boys and girls, it wasn't the Web, it was [inaudible], it was too bloody hard.

I'm curious about this though... Beta 2 was coming up and Beta 1 was already at least somewhat feature complete from what I know; correct me if I'm wrong. Maybe practical feasibility was the problem, but I don't know if it's really technical feasibility like this guy says.