Spotted in Making it stick, There seem to be some issues floating around with the new Microsoft WinFS. At "Making it stick" he accuses the WinFS team of being stuck in the past. I don't fully agree with this, but I do have my own sticking points with winfs.

The Network is the Computer seems to be something that the Microsoft team has actually tackled in an impressive manner. They only briefly touch on it in their live video, but they have a full synchronisation system built in to WinFS. They touch on the fact that I can cache a local copy of a networked document, edit it back home, come back in to work and have it synchronise the changes correctly. They even seemed to hint that it managed ownership of these things correctly.

This seems to me to be a huge step forward - if it's true. So let me put on my cynical hat for a second and complain that this is just another step by Microsoft to win the database war.

Why is that you ask? Well, one way to win the database war is to make sure every one has a database and that programs are built to work with that database only. How do you do that? you make all the most common programs people use use that database with specific schema's that hold information other apps will be interested in.

Oh and what about network accessible databases? Obviously they've got that too. SqlServer already does that and now you can access a database as if it were a file system (*cough* how long ago did Oracle have this? - if only they'd thought to integrate it seamlessly in to Windows Explorer *snicker* and bundle it with upgrades and OS's *double snicker*).

So it's no longer a matter of "Why would you use Oracle" it becomes a matter of "All the relevant data is already in SqlServer". Very tricky - exactly the kind of technology I'd expect from Microsoft.

So, taking off my cynical hat for a second and putting on my futurists hat.. where has the WinFS team fallen down. Like I said, the synchronisation ideas are excellent, but what about the whole concept itself. A database instead of a file system?

If you listen closely, you'll heard that WinFS actually runs on top of NTFS and when you copy a file in to a WinFS file system, it actually puts the real data in to the NTFS file system with various metadata stored in the database.

Hello rewrites! Every app that wants to take advantage of this new paradigm will have to be rewritten. You heard me. Every one of them. Why? Because you cannot treat it like a file system and get access to all the other records that you want to deal with. It's simply not possible. You have to treat it as a database. Once you do that, you can enjoy their relationship constructs.

So if we have to rewrite our applications to be DB apps instead of FS apps, what will the cost be? Well, the biggest risk and cost is that this is not the end-game technology that it's being advertised as.

And it's not.

What they should have been working on these past few years is making a Topic Map like file system to allow auto-discoverable relationships and associations between files that do not require all applications to be rebuilt every time Microsoft comes up with a new wizbang relationship that they're going to put in to Office.

Topic Maps have the unique ability to let a system grow forever while still represent real semantic value and still being synchronisable. Yep, you can even put topic data in to a relational database - think Microsoft will? Hell no.

So, what's to stop someone else from doing it? I'll tell you:

• Word won't use your topic map based relationships
• Neither will Windows Explorer
• How will you get others to adopt your standard exactly?

So the more things change the more they stay the same. The only relationship types that will really matter in WinFS are the ones Microsoft uses in Office. Big surprise there. And to rub salt in to the wound, they didn't take the next important leap in computing and implement Topic Maps as the standard relationship construct.