Microsoft's R&D chief: the people problem with innovation

Correction

Rick Rashid is the head of the Microsoft Research organization, not the head of Microsoft R&D, and the budget of $9billion is the budget for the whole R&D (not Microsoft Research that Rashid heads).

Heads in the clouds

The problem with cloud computing or whatever the evangelists like to call it, is trust.

And that is the significant hurdle they have to overcome, even if their products are perfect, which looking at the past, they may not be, the users, whoever they are must have absolute faith in the provider.

And I for one certainly dont have anywhere near that kind of faith to trust Microsoft by a long way.

tl;dr

hindsight ftw

R&D

With billions spend on R&D by the brightest and best computer scientists in the world, Microsoft are still incapable of making an OS which doesn't overwrite an existing OS on the same disk.

Anyone can have a great idea.

Games companies get inundated with them every day from hopeful third parties. The trouble is that in order to make a great product, everything depends upon the details of implementation - those minute-by-minute decisions which lead to a quality product. The same holds true for music, movies, books, cars, gadgets and toys.

Great idea + crappy implementation = useless end result

Crappy idea + great implementation = potentially great product

Great idea + great implemetation = killer app

Timing is everything

Sun actually tried to market cloud computing in the 90s, but it failed because it makes a telephone wire a single point of failure for your business. That's still true. If this guy "had the idea" a few years ago, then perhaps in another few years time he will learn the lesson that Sun learned (or did they?) about ten years ago.

Had mobes existed 30 years ago, the match between Windows 1.0 (very small, no proper protection, only reliable if you control all the apps) and that platform would have been obvious. By the time such devices were widespread, Windows was far too large and had to be stripped down to make CE.

What?

"Rick Rashid, leader of Microsoft's vast, multi-billion dollar R&D operation, said he could foresee cloud computing some years back.

The challenge as a technologist, though, has been in anticipating the finer details of how the cloud and its related technologies - the data center, replication, and synchronization - will be adopted by people and organizations."

I predicted it even before that in the midst of a drunken slumber and I couldn't foresee the finer details either. Bollocks.

I'll be sure to post in a few years time about some of the other current tech-buzz happenings that I foresore but wasn't sure about the fine print thereof. Tit.

Cloud computing is easy in theory

its near impossible when you have a major player ignoring all common standards and the patenting everything in sight in a desperate attempt to own everything rather than actually implement.

Mobile phones did exist in the early 80s

Motorola DynaTAC was the first US approved cell phone in 1983.

Innovation

As far as touch screen user interfaces, I believe Dave Kurlander's work that later led to their "Surface" product had a lot of the ideas you see on the iPhone.

It's a tired cliche that Microsoft never innovates, repeated by people who don't really know anything about technology, but hate big corporations in general. Microsoft's NT operating system was lightyears ahead of UNIX in 1990, and Linux is just now catching up to it in terms of feature set. Windows 95 had a pretty good UI for its day, many of its ideas sneaking their way back into Apple's system later (clearly a lot of give and take between those two companies). OLE Automation was a major innovation that allows extensible interfaces on applications, something the Linux community tried and botched with the Bonobo fiasco. Windows was far more modular than UNIX, with dynamic libraries and object interfaces (COM), and device driver interfaces -- all ideas copied years later in Linux. If you start to look at multicore hardware, NT is still leading in terms of its Amdahl-law characteristics (Linux levels of at 4 processors, NT at 8 and Microsofts new red dog is rumored to reach 128 processors before reaching assymptotic performance). The layman tends to look at superficial things, like the widgets in Vista and Mac, oh who had those first (well, Konfabulator and other 3rd party software did that long ago). They don't understand what goes on under the hood, and under the hood the NT kernel is still superior technology to the Mac or Linux operating systems.

@Don Mitchell

It's a tired cliche that Microsoft never innovates, repeated by people who don't really know anything about technology

And that's a tired straw man argument put up either by Microsoft sock puppets or people who are deeply ignorant of the *history* of the technology they are talking about. On the off chance you are the later prepare to learn.

Microsoft "innovate" by buying up companies,hiring staff *after* they have made innovations or buying patents to enforce then getting saturation PR to describe it as ground breaking.. This is not innovation as most people understand it.

"Microsoft's NT operating system was light years ahead of UNIX in 1990,"

Debatable. What features did you have in mind?

What's not debatable was the Director,Windows NT Development, David Cutler(and most if not all of his team) came from DEC where he designed VMS. The most compatible part of Windows NT with Windows 9x was calling it Windows.

"and Linux is just now catching up to it in terms of feature set."

Which ones did you have in mind?

"Windows 95 had a pretty good UI for its day, many of its ideas sneaking their way back into Apple's system later."

Neatly airbrushing out O/S2 and the long term work done by Xerox, which pre-dated all Apple & Microsoft work in this area. Windows keyboard short cuts are straight out of the IBM Common User Access guidelines BTW.

"OLE Automation was a major innovation that allows extensible interfaces on applications."

Setting inter-process communication standards (which OLE, ActiveX or whatever your going to call it are) is a lot easier when you have a proprietary OS architecture (Amiga Rexx anyone) or a virtual monopoly on a very limited number of hardware architectures (x86 compatible and what others these days?). The IBM "Message Queue" approach is cross hardware platforms.

Corba is the platform and OS neutral standard for this but I don't know how widely it is used. The others were and are real cross app scripting and control options around when OLE came out. Consensus in a monopoly is easy. Its tougher in a community and tougher still to explain why having it is a *good* thing. I note the common lack of qualification. This only applies to non Windows with significantly greater difficulty.

"Windows was far more modular than UNIX, with dynamic libraries and object interfaces (COM), and device driver interfaces -- all ideas copied years later in Linux. "

Funny, part of Microsoft's claimed reason for resisting the un-bundling of IE was that explorer code was so entwined with the core modules it was impossible to remove.

I note your now moving the goalposts from Linux to Unix in general. Unix always had the idea of high level "block" or "character"mode I/O devices rather than a whole different API for each device. IIUC Linux made them installable without a full scale kernel re-compile. Something (I think) even the original Unix developers noted under "What we would have done differently." Linux might have gotten DDL from Windows. Or OS400, an OS (and apps) originally designed to run as a large set of DDL's to reduce disk and memory footprints and allow running form the disk directly. Hard core eXecute-In-Place for the palmtop developers reading this. I don't know enough other architectures in depth to comment if there could have been other sources. Burroughs B5000 for example?

"If you start to look at multi core hardware, NT is still leading in terms of its Amdahl-law characteristics (Linux levels of at 4 processors, NT at 8 and Microsoft's new red dog is rumoured to reach 128 processors before reaching asymptotic performance). "

NT (do you not mean an NT derived OS like Vista) would still be behind VMS (same architect) which IIRC topped out at a 64 processor cluster. And Amdahl's law warns that the speed-up hits the buffers when any part of the program cannot be parallelised. You'll need an app which is <0.78% serial to actually exploit that.

"They don't understand what goes on under the hood, and under the hood the NT kernel is still superior technology to the Mac or Linux operating systems"

I'm not sure how much you understand either. “Superiority” is debatable. Expensive, insecure,proprietary, bloated and designed to force customer lock-in. definitely

To anyone who grew up using Microsoft kit only it can seem very impressive. To those of us who know of the wider IT world its rather less so. Apple's system later."

Neatly airbrushing out O/S2 and the long term work done by Xerox, which pre-dated all Apple & Microsoft work in this area. Windows keyboard short cuts are straight out of the IBM Common User Access guidelines BTW.

"OLE Automation was a major innovation that allows extensible interfaces on applications."

Setting inter-process communication standards (which OLE, ActiveX or whatever your going to call it are) is a lot easier when you have a proprietary OS architecture (Amiga Rexx anyone) or a virtual monopoly on a very limited number of hardware architectures (x86 compatible and what others these days?). The IBM "Message Queue" approach is cross platform.

Corba is the platform and OS neutral standard for this but I don't know how widely it is used. The others were and are real cross app scripting and control options around when OLE came out. Consensus in a monopoly is easy. Its tougher in a community and tougher to explain why having it is a *good* thing.

"Windows was far more modular than UNIX, with dynamic libraries and object interfaces (COM), and device driver interfaces -- all ideas copied years later in Linux. "

Funny, part of Microsoft's claimed reason for resisting the un-bundling of IE was that explorer code was so entwined with the core modules it was impossible to remove.

I note your now moving the goalposts from Linux to Unix in general. Unix always had the idea of high level "block" or "character"mode I/O devices. IIUC Linux made them installable without a full scale kernel re-compile. Something (I think) the original Unix developers noted under "What we would have done differently" a *long* time ago.

Linux might have gotten DDL from Windows. Or OS400, an OS (and apps) originally designed to run as a large set of DDL's to reduce disk and memory footprints and allow running form the disk directly. Hard core eXecute-In-Place for the palmtop developers reading this. I don't know enough other architectures in depth to comment if there could have been other sources.

"If you start to look at multi core hardware, NT is still leading in terms of its Amdahl-law characteristics (Linux levels of at 4 processors, NT at 8 and Microsoft's new red dog is rumoured to reach 128 processors before reaching asymptotic performance). "

NT (do you not mean an NT derived OS like Vista) would still be behind VMS (same architect) which IIRC topped out at a 64 processor cluster. And Amdahl's law warns that the speed-up hits the buffers when any part of the program cannot be parellelised.

"They don't understand what goes on under the hood, and under the hood the NT kernel is still superior technology to the Mac or Linux operating systems"

Again debatable. Expensive, insecure,proprietary, bloated and designed to force customer lock-in. definitely

"It's a tired cliche.." "repeated by people who don't really know anything about technology"

I know a little of the history of this technology. I have described it on the off chance you merely ignorant of it. To anyone who grew up using Microsoft kit only it can seem impressive. To those of us who know of the wider IT world its rather less so.

@ Don Mitchell

So, let's be clear about this:

1. You're saying that no Unix had dynamic libraries before Windows?

2. You're saying that no Unix had device drivers before Windows?

3. You're saying that no Unix had object interfaces before Windows?

Perhaps I have misunderstood you, in some strange way, but at the moment your grasp of "what goes on under the hood" (or "bonnet", as we say in English) seems to be roughly on a par with the staggering percipience of Mr Rashid, who managed to remain unaware of the existence of mobile phones years after they were invented (and quite possibly after I;d made my first call on one).

John.

re: Avie Tevanian?

If there's anyone who can take credit for building Mach its the PI. Obviously, building an OS takes a lot of work. There's plenty of credit to go around.

But the PI had to come up with the ideas, find the money, justify the research, prioritize things, etc. Did the architect build the house, or was it the guy who hammered the nails? (And think about the people in between.) They can all rightly say they built the house.

@John Smith

"So the answer is no."

Quite sure? Final answer?

I am not sure what year any of the various OSsen that go under the name of "Windows" introduced dynamic libraries, nor when the relevant POSIX standard was defined, but I'm pretty dam' sure that Sun Unix at least had the beasties in, ooh, lessee, 1989.

Now I've stirred the memory pot, I wish I could remember if you could use external .atr files in Simula on any Unix. I only played with them on VMS, unfortunately.

Good to know some people remember Multics; I tend to doubt that many microsofties were ever Multics refugees, though. Remember that Win 3.1 still hadn't quite managed the leap to multi-programming, as Bill thought it unnecessary on a PC, hence the dreadful nonsense of TSRs.

Good to see OS/2 remembered, too. I can remember a time when the fastest and safest way to run Windows applications was to run them under OS/2 Warp.

But, as ever, the Microsoft fanboys try to re-write history -- possibly because they are simply ignorant of it in the first place.

All the best,

John.

@ John Smith

""Ken Ritchie" is a Freudian slip. That should have been Ken Thompson."

Don't worry, I didn't spot it. After all, who bothers with surnames for Ken, Brian and Dennis?

All the best,

John.