About | Buy Stuff | News | Products | Rants | Search | Security
Home » Resources » Rants

SWANT

Week of March 15, 2002
Switch to Linux now so you can laugh at these clowns when they finally fall on their faces.

You would prefer ONE CHEESE? And ONE CHEESEBURGER to put it on?
 - Luc Teyssier

Most people don't believe Microsoft really have a department called 'Secret Weapons & New Technology'. 'Nobody is that immature' is what they mean. But they do. And that's exactly what the department is called. 'Secret Weapons & New Technology'. And the head of SWANT is just what you'd expect.

CNET had a story earlier this month on Microsoft's supposed push for a 'new Cairo'. You all remember Cairo, right? That operating system that became NT4 instead? And that Bill Gates Himself later claimed was never an operating system, but just a star to shoot for? Yeah sure. Cairo really existed, and anyone unfortunate enough to have seen the plans for this monstrosity must be grateful today that the geniuses in Redmond weren't able to pull it off. Which is begging the issue, as something as dumb as Cairo would have been impossible anyway.

Consider their planned Object File System or OFS: The idea for OFS was based on their OLE2 streams and storages, and if you are familiar with OLE2 streams and storages, then you are probably running for your flight bag right now. And if you are not familiar with these other instances of typical Redmond brilliance, then why don't we take a moment and bring everyone up to speed?

How best to begin? Ok - have you seen how these Office applications can offer a 'fast save' on files? And you are all aware of how Word and what-not can all save Excel clips and pictures and what-not all in the same document file?

What's going on here is the first extrapolation of DDE - Dynamic Data Exchange. On Windows 3.1x the new OLE - Object Linking and Embedding - was run by using DDE as a sort of 'carrier wave'. DDE in turn is just a kind of 'dynamic clipboard' where different apps can communicate by sending and receiving data over globally accessible areas (like the clipboard).

Which leaves us where? Well doing all this stuff is buggy enough - trying to save it to disk is about ten times more buggy. It takes time - LOTS of time. And how in hell do you structure it all on disk? What if you have containers such as Word documents for objects such as pictures but also for OTHER containers such as MORE Word documents that in turn contain MORE Word documents and so on?

You have storages and streams is what you have - this is how you (try to) solve that problem. Basically what you get is 'a file system within a file' (don't run for the flight bag yet please). Within one single file you establish 'storages' which are like a kind of directory, and within these storages you place the data - here called 'streams' because by not calling it 'data' you can get people really confused, and it's a really groovy word too.

Ok, so what we got is storages - directories - and streams - files. So this takes quite a while in realtime to construct if you want to save this junk to disk.

So what the Office applications do with their 'fast saves' is they ignore FRAGMENTATION inside these odious creatures; they just find the storages they want and store the streams there, and the disk waste be damned.

Take for example a Word document with several 'storages' where one storage has a text segment of - let us say - exactly 4KB - 4096 bytes. So you open this document and edit it - and when you want to save it to disk, the text segment is now 4KB + 1 - 4097 bytes! What do you do now?

Well if the file was stored compactly, you are going to have to find new disk space for that stream. Now here's the punch line: on a 'fast save' you ignore the previous 4KB segment - you just leave it in there! But it is no longer referenced, so it's never used, and aside from file fragmentation you won't be harmed - at least in theory.

Ok, so basically we're looking at a technology where there are a lot of things that are suspect, but right now we'd like to address the 'fragmentation' that takes place, because this can be a very serious problem further down the road, right?

Right. And Craig Brockschmidt, official head of the Microsoft OLE2 team (even though he didn't program in C++ and hated it) wrote THE BOOK about OLE2 - Inside OLE2, the book with the famous first sentence of introduction:

This is a book about fish.

Craig might not have known what he was doing with these OLE2 guys. He might have lost his sailing charts. Whatever. He did write 'the' book, and he did provide programs for it as well.

One of the programs was an attempt to DEFRAGMENT an OLE2 'storage and stream' file. IT NEVER WORKED.

And actually, when you think about it, the idea, which boggles the mind, of a 'file system within a file' is so staggering that your instincts would tell you 'defragging is for all practical purposes impossible or at least improbable so why not trash this stupid idea and try to move on to something more intelligent and with more promise?'

Why not indeed?

But never count the idiots in Redmond out - OFS is simply a global extrapolation of OLE2 storages and streams.

Think about it, dear computer users: Microsoft in the future wants your disks to have ONE FILE and have EVERYTHING within this single file. Now - honestly - how do you think that is going to work?

Switch to Linux now so you can laugh at these clowns when they finally fall on their faces.


Like other secret weapons in the technology business, Microsoft's vision of a unified data store in its Windows operating system has been the source of speculation - and the butt of industry jokes - for a decade. Now, Microsoft wants to resurrect its plan as part of the next version of Windows. But we've heard this all before.

1992: Jim Allchin details a vision for a new version of Windows, code-named Cairo, which would include a revamped user interface and a new data store, called the Object File System (OFS), for storing document files, spreadsheets, multimedia files and other information. The goal is to enable searching not only by file name, but by file content. Cairo would also include a new directory service. The operating system is expected to debut in 1994.

May 1994: Allchin tells the press that Microsoft has reassigned programmers to work on the upcoming Windows 95 desktop operating system, delaying the shipment of Cairo until late 1995.

November 1994: Microsoft says Cairo will slip again, this time until 'sometime in 1996' then-executive Vice President Mike Maples told Computerworld.

August 1995: Microsoft ships Windows 95, which includes the revamped user interface originally slated for Cairo. Microsoft says the first test version of Cairo will debut in late 1996, pushing the release date to 1997.

December 1995: Gates launches his 'Internet initiative' to make Web technologies a top priority at Microsoft. Work on Cairo is reassessed.

1996: Bill Gates says Cairo is a 'vision,' not a product, leading many to believe the project is running into a roadblock. Gates tells Computerworld: 'Cairo is a futuristic system. It's something we're working on.'

March 1996: Microsoft says work on the OFS for a future release of Windows NT has been abandoned in favor of a file system based on its Exchange Server.

August 1996: Gates says some Cairo technologies, including indexing capabilities and a distributed version of Microsoft's Component Object Model software, will be included in Windows NT 4.0, and directory features will be part of NT 5.0. The product jockeying causes Goldman Sachs analyst Rick Sherlund to say: 'I think Cairo has lost its definition.'

1997: Rumors circulate that the OFS will debut by year's end. Gates tells PC Magazine: 'The only thing of (the Cairo) vision that's not in the marketplace is the file system and directory. And so later this year that'll go into beta testing.' NT 5.0 enters beta testing, but the file system is never seen.

October 1998: Microsoft says it will rename Windows NT 5.0 to be called Windows 2000. The operating system will include Active Directory, technology originally slated for Cairo.

February 2000: Windows 2000 debuts after several delays.

2001: Allchin says he has 'not given up' on the OFS concept.

January 2002: CEO Steve Ballmer says, 'We want to evolve our storage system.' Allchin says the OFS unified storage concept is again in development and will be a major part of Longhorn, the code name for the next version of Windows.

About | Buy | News | Products | Rants | Search | Security
Copyright © Radsoft. All rights reserved.