Dr. Jerry Pournelle

Email Me


Why not subscribe now?

Chaos Manor Subscribe Now



Useful Link(s)...

JerryPournelle.com


Hosting by
  Bluehost


Powered by Apache

Computing At Chaos Manor

Mailbag for September 18, 2006
Jerry Pournelle jerryp@jerrypournelle.com
www.jerrypournelle.com
Copyright 2006 Jerry E. Pournelle, Ph.D.

September 18, 2006

There was a lot of interesting mail on many subjects this week. Begin with a reference to the good news that my medical problems are related to arthritis in my neck, rather than the something a great deal more serious that we were afraid of. It turns out my posture is important.

Subject: Neck Brace Tests

Dr P.,

Is a neck brace nature's way of telling you that it is time to review the state of the art in voice recognition and transcription to word processing?

Jim w.

I have often experimented with the Dragon Naturally Speaking system, and it works pretty well. If I were doing mostly photo-journalism, and tour talks, I would use it in an instant; I was able, for example, to do commented photo tours in Japan by dictating into a portable device, and with the wonderful little Griffin device that turns an iPod into a highly portable speech recorder. But, it turns out, writing fiction and non-fiction is not so simply done, at least by me. It turns out that while I can stand up in front of an audience and make a half hour speech that is coherent and organized from notes, I don't seem to be able to write these columns, and fiction scenes, that way. It may be that I will have to learn how to do that, so stand by.

There is a new version of Dragon out, and I need to find it. And on the same subject:

Subject: Bifocals?

Jerry,

Please disregard if you've already thought of this...but...do you by chance where bifocals? If so, have you investigated those as the source of your "pain in the neck"?

I vaguely remember your discussing this in BYTE back in the 80's(?), but I've only recently found you again so I'm not sure if you've revisited this.

As a professional "sit at the computer monitor all day" geek, I found that while wearing bifocals I usually developed either a severe pain in the shoulders and neck, or a severe sinus-like headache. It seems to be related to the awkward position I hold my head in order to look through the "closeup" lens'.

My solution was a pair of computer glasses...not bifocals, but with the correction adjusted for close up work. That way, I don't have to tilt my head.

(As it happens, I'm far-sighted in one eye and near-sighted in the other. As I've aged, my brain has learned to switch between them so I no longer need glasses at all even though my overall vision has deteriorated).

Possible?

Don Etling

I believe I invented computer glasses. That is, in 1980 I asked my optometrist to make me a pair of spectacles with a focal length of 28 inches, which was the distance I sat from the screen of Ezekiel, my CompuPro Z-80 system that is now on display in the Smithsonian (Museum of American History, Hall of Communications and Computing; he's in there standing on a stack of BYTE magazine alongside an Altair and an IMSAI paddle switch system). I wrote that up in my BYTE column, and I am certain I never saw any reference to computer glasses before I did that.

So, yes, I can sit a good distance from the screen, and I adjust my head position often, and I don't have to tilt my head to use the bi-focal element, which is why I invented computer glasses in the first place. I do have to be careful not to try to drive while wearing those.


Subject: How Should We Use this Power?

Dr. Pournelle,

I've been thinking a great deal about your column regarding multiple core processors ("An Era of Computing Plenty," 31 July 2006) and following the recent debates about Windows Vista. Mind you, I'm a Mac user and have been for nearly a decade now. However, I think that if anyone wants to find something useful to do with all this computing power, that person should forego eye candy for something really useful: True, honest-to-goodness voice recognition.

It may seem strange coming from a Mac user, but I don't think much any more about the interface. The interface I have on OS X Tiger is more than sufficient for my needs, and it works well on my G4 14" iBook (with a processor that runs less than 1 GHz, if you can believe that). However, if someone could develop voice recognition so I could forget the interface and instead communicate with my computer, I'd buy as much computer as I could afford. Heck, I'd probably try to spring for something I could barely afford, and that's on a country pastor's salary supplemented with adjunct teaching at our local community college.

Why didn't Microsoft go with this instead of adding all these CPU-hogging interface gadgets? The interface works. Give us communication instead! Let us tell the computer what to do instead of making it do what we need it to do. I'd suggest that the first OS that accomplishes this will quickly overwhelm any competition.

Then again, maybe Redmond is waiting for Cupertino to pave the way....

Hope your sinuses are better. We're suffering from them here in the South, too. It has been a hot and dry summer.

JA

John Alexander

"The Christian faith has not been tried and found wanting. It has been found difficult and left untried." --G.K. Chesterton

Dragon Naturally Speaking has gone a long way toward genuine speech recognition. It is certainly one important use for our new computing power.


Subject: multiple core processing

Jerry, you may recall that about a month or so ago I wrote you, and suggested that the future of computing would be the evolution of parallel processing in home computers. I suggested that over the next ten years we would see a proliferation of multiple cores, larger caches and more independently addressable memory. I was wrong; it's here already: (link here).

The article examines the performance of Intel's new quad core processor; briefly, it slams AMD to the mat. There's also a short discussion regarding AMD's response, due out very shortly; I expect that AMD will be working double overtime to ensure that they recapture the performance crown. There's serious money at stake; AMD is fighting for more market share, and - having recently announced 10,000 layoffs - Intel is fighting for its life. Given the vigorous expansion of parallel computing, I think we're going to see a real trend in software products to take advantage of the new hardware. Look for parallel processing optimized operating systems, and languages specifically designed with the new capabilities in mind. I also don't expect to see much more evolution in clock rates; instead, I think we'll see progress in using power more efficiently, in order to generate less waste heat. This translates into better laptop battery life, and quieter desktop operation.

And so the next bold frontier is parallel processing; I now think that we can expect to see hundreds of mini cores on a chip before long. We're really at the bare, fumbling beginnings of parallel processing; the chips we build five years from now will be vastly more sophisticated than anything we can currently achieve. There's a lot of 'growing room' here. It amazes me how silicon is constantly evolving to more and more closely model the human brain. True artificial intelligence is a lot closer than we might think; Ray Kurzweil thinks around 2029, but he may be too conservative. It could happen sooner.

And if the artificial intelligence we create believes itself to be self aware... what does that say about our own belief in God and an afterlife?

Regards, Charlie

I am not much worried about an omniscient and omnipotent computer taking over the duties of the Almighty. I do note that decades ago, Freeman Dyson speculated that we might achieve the equivalent of personal immortality by downloading our personalities from meat-based hardware into sufficiently powerful computers. I believe I called that chapter in A Step Farther Out "The Integrals of Immortality."

Chaos Manor Associate Eric Pobirs comments:

Part of the security presentation [at DefCon] touched on another recent column issue, that of 'what'll we do with all of this power?' Being able to do away with paging thanks to massive amounts of RAM would be sweet, and if it closes an avenue of attack, all the better. It would be interesting to compare the power usage for very large (by current desktop standards) memory installations vs. using the hard drive for virtual memory. If MRAM densities got high enough and cheap enough the difference would be huge but I wonder where it is with current RAM types. Could a 'greener' PC also be better in other ways?

Eric


Subject: Upgrade from Word 2003

"My guess is that when you upgrade from Word 2003 to Word 2007 you will lose about a week of productivity over the course of the next six months, as experienced Word users tear their hair out trying to discover how to do things they used to do instinctively."

When I switched from Word 2000 to OpenOffice Writer as my primary word processing program I had a similar, although much less painful experience. I find OpenOffice 2.0 does everything I need, supports MS Office proprietary formats reasonably well, but also offers a standards based document format (opendoc) as its default.

I think the real question is how many people are going to pay hard earned money to learn a new word processing program when they can do the same thing for free from openoffice.org. I've had several conversations with people in the last several months that were buying computers and I suggested they try OpenOffice before spending money on a new version of MS Office. So far I think my conversion rate is about 50% even before the effects of the disruptive changes in Office 2007 kick in. I don't think this bodes well for MS Office.

Scott Kitterman

As I have said, I have so far seen no compelling reason to convert from Office 2003 to Office 2007. I do like 2007. It has some significant bugs which I presume they will fix; the question is, if you have to pay money for the upgrade, why do it? I have no definitive answer to that.


Subject: Vista Upgrade

Hi Jerry,

Just read this week's column, and had some thoughts. I believe that a number of factors are about to converge (singularity?) that will provide an opportunity for a major shift in the desktop market.

1) XP and existing machines are perfectly adequate for 99% of all home and corporate applications.

2) Vista requires significant cost, both in hardware and installation, to deploy

3) MacOS now has the ability to run WinTel applications nearly seamlessly

4) Linux is viable for 99% of all home and corporate desktops

5) OpenOffice is viable for 99% of all home and corporate uses

6) Most corporate applications have, or are, migrating to a thin-client (browser) structure

7) The thick client applications run very well on Citrix farms

8) MS Office's redesign requires as much productivity loss as does switching to an alternative. Microsoft is taking a huge gamble when they reduce the barriers to competition - if it takes as long to move to OpenOffice as it does to Office 2007, why not choose the cheaper solution?

9) Linux runs just fine on existing machines

10) Apple's price/performance premium is at a historical low (and the OS is 1/2 to 1/3 the cost of Vista).

Combined, if I were a CIO, I'd be asking two questions "Why should we upgrade?" and "What should we upgrade to?". Eventually the former is moot - Microsoft will only supply XP security patches for so long. For the second it's much less clear, but here are some thoughts and what I'm recommending:

1) Home users should go to MacOS and OpenOffice. I know Linux advocates say that it's ready for Aunt Minnie, but if something goes wrong, a novice user can always walk into a Mac Store and talk to a Genius, or call support and talk to someone that speaks English. Linux support is spotty at best, and I can only do so much tech support in my spare time. If they need to run a few Windows apps, run parallels and XP. If the XP app never talks to the Internet, you can pretty much do that until the end of time and be safe.

2) Corporate users with small .Net or thick windows client applications: Linux Desktops, OpenOffice, and Citrix for the handful of applications. Alternatively, MacOS, but the advantage is Linux, because it can be installed on existing hardware at minimal cost and improved performance. Both are UNIX underneath, so there's a common technology for the gearheads.

3) Corporate with large .Net or thick windows client applications: First choice is to stick with XP and begin an architectural migration away from client-side logic (except for certain apps like video processing and CAD). Citrix is an alternative, but for some Vista will actually be cost-effective.

4) Small businesses: This one is tough - if there's a techie on staff, go to Linux. If they only do basic applications, go to MacOS X. Unfortunately, for most they'll need some windows specific apps, so they're stuck with Vista.

I believe that the bulk of the current world falls into categories 1 or 2. The convergence of capabilities and costs make options viable that never were until now. We had a similar situation with OS/2 way back when, but IBM couldn't market free beer to construction workers, so that opportunity was missed. The question is: will Apple and the Linux vendors learn from the OS/2 failure, or will more great beer go down the drain?

Cheers,

Doug

"Do something you like. Forget about the pay, for Christ's sakes. Regulate your style of living to fit your income. Just have fun in your job, that's the main thing."
~ General Chuck Yeager

I am strongly considering switching to an Intel Mac and using that for testing both Vista and XP applications; we will be looking into that in future. I will have much more on that in the column.

Peter Glaskowsky comments:

> 1) XP and existing machines are perfectly adequate for 99% of all home and corporate applications.

Sure, it always looks that way just before an OS upgrade. But a year later, we'll look around and realize that the best applications-- the new ones we really want to run-- only work on the new OS, or at least work best on the new OS. So we buy a new machine, get the new OS, and learn to deal with it.

> 4) Linux is viable for 99% of all home and corporate desktops
>
> 5) OpenOffice is viable for 99% of all home and corporate uses

Except Linux and OpenOffice actually suck. Somehow, Linux and OpenOffice enthusiasts can't see that. They assume that 99% of the world is like them, but that's not how reality works. People who do the atypical things are not like 99% of the world. They're atypical. The rest of the world looks at Linux and sees something that is years behind Windows or Mac OS in visual appeal and ease of use. Sure, Linux is more hackable. Most people don't merely not care, they actually don't want that. Big surprise.

Some of our people at the office use Linux-only apps, so they have to use Linux, which means they have to use OpenOffice for documentation-- which means that I have to use OpenOffice to work with their documentation. I dread getting one of these files to work on. OpenOffice is slow, awkward, ugly, violates good user-interface principles generally and Mac GUI rules specifically, and lacks features I use regularly in Microsoft Office.

> 6) Most corporate applications have, or are, migrating to a thin-client (browser) structure

Thin-client systems violate a rule I've always considered to be a generalization of one of Pournelle's Laws-- local processor performance tends to increase. People only give up their own computers when you pry their cold, dead fingers from them. Thin-client systems are entirely incompatible with high performance, mobility, and-- perhaps surprisingly-- low cost.

A thin client still needs a processor somewhere. Processor equivalents are MORE expensive in servers and in the server room. Real estate is more expensive in the server room. Power is more expensive in the server room. Maintenance and upgrades are MUCH more expensive in the server room.

The costs of systems with thin clients plus centralized processing have always been higher than the costs of PCs on a LAN. They always will be.

Pournelle's Law is One User, at least one CPU. A corollary is that no one, deep down, wants his files out in an Internet Cloud, or even on a corporate server. We all want our stuff on one own machines. Another of Pournelle's Laws is that no one wants to share CPU cycles with anyone, including himself.


Subject: Vista 64 vs 32 bit

Jerry,

I have been running Win XP x64 for over a year now and here is what I have found.

If the application uses a 16 bit installer you are out of luck. 32 bit applications with 32 bit installers work just fine.

Performance seems to be better overall than running the 32 bit version of XP.

The major problem that I have had is finding 64 bit drivers for some of the hardware that I would like to use with the system. My Hauppauge WinTV PVR2 USB being the primary problem.

NVidia and ATI video drivers have not been a problem and I was forced to switch from Norton Anti-Virus to Avast! I should thank Symantec for the lack of a 64 bit version outside of the Corporate 10 pack. Avast has been reliable and FAST with almost daily updates to the database. Best of all it is FREE to Home users.

I use a Highpoint Raid card on this system and there has been a 64 bit driver for it since I purchased it last year in February. The only problem that I experienced was in moving from the x64 beta to the shipping version. I had to modify the preamble of the .inf file in order for the Highpoint driver to load. (Not a big deal after trolling the Microsoft support database and learning what needed to be done.) Why MS decided to make changes to .inf files between the final beta and the shipping version eludes me. Perhaps some anal retentive tech decided that the change was necessary.

I think that you would be doing your readers a service by going with the 64bit version. It also might prove to be more interesting for you for you too. (I don't mean interesting in the "may you live in 'interesting' times" sort of way.

Bob Holmes

Thanks. I do intend to test out Vista 64, but first I want to get used to 32-bit. We'll get to 64-bit soon enough.

David Em has found that there are some severe driver problems with Maya, a 64-bit application; he can't get the dongle to work properly.


Subject: Windows on Intel Macs

Jerry,

Used to hang out at Chaos Manor quite a bit, and am glad I checked back in recently. A long time Mac Nut, I'm also glad to see that you're getting interested in them. I think a lot of us have been surprised and pleased that the Intel transition has been as smooth as it appears to have been. The new desktops are simply awesome, but a lot of us are waiting to see if a dual quad core Xeon box with be announced at MacWorld/SF in January or shortly after, since (as I understand it) Intel is scheduled to release quad core Xeons the last quarter of this year. In any case it never hurts to wait awhile and let any "undocumented features" that pop up with these 1st gen boxes get fixed.

The real icing on the cake has been Apple's announcement of Boot Camp, their solution to installing XP and booting Windows apps on an Intel Mac and Parallels Desktop. However, while I could be wrong, I think the Windows installations on a Mac are just as prone to all the virii and malware as a Windows OS running on native Windows hardware since it's the software that's vulnerable in the first place. We'll all find out soon, I'm sure.

BTW, I noticed your mention of Bill Gates and his role in the early days of personal computing. If you haven't read "Dealers of Lightning" by Michael Hiltzik, I highly recommend it. It's the story of Xerox PARC and focuses on the people and personalities with the technologies developed at PARC being the framework on which the stories are hung. I think an interesting and well written book.

Take care,

Ronnie Day

Boot Camp is interesting, but I find Parallel more exciting, and almost certainly more secure: a Windows virus should starve in a Mac OS environment. Of course we have to test out performance. I want to see World of Warcraft in Vista on an Intel Mac...

Regarding waiting for the quad core systems, Peter Glaskowsky notes:

And then maybe he'll wait for a dual 8-core machine, and when THAT shows up, he'll say that dual 16-core machines can't be very far away...

C'mon, the Mac Pro has four 3.0-GHz CPU cores. It's faster in practical terms than any machine Cray Research ever shipped. There won't be any big step-function improvements in performance in the next three years. Anyone who can use one should buy one.

. png

And my only real dilemma is whether to get the newest Intel Mac laptop or go for the really big desktop hardware. I sure find them interesting.


Regarding 802.11 "n" network incompatibilities:

It was my understanding that "Draft-N" hardware was the hardware responsible for RF interference with 802.11b/g networks whereas "Pre-N" hardware was more amenable to 802.11b/g network gear. More specifically, equipment using the second generation AirGo chipset (which the Belkin Pre-N uses) seems to work well. Broadcomm and AirGo's third generation draft-N chip and Marvel's Draft-N chipsets seem to be less then favorable to other hardware then itself.

The reason the Draft-N hardware is obliterating 802.11b/g networks is due to the status of the 802.11n proposal at the time these chips were all drafted. AirGo had been trying to push for spectral efficiency in the 802.11n standard, using only one 20Mhz wide channel for traffic. However Broadcomm and Marvel used a 40MHz wide signal for their Draft-N chipsets which allowed them to regain the speed crown from the 2nd generation AirGo Pre-N spec based chipset. Once AirGo saw that the IEEE 802.11n standards body wasn't going to rethink spectrum efficiency they basically cranked all the specs to 11 on the 3rd generation Draft-N based chipset.

That all being said, any new wireless network equipment I setup now I force to provide boring single channel 54Mbps 802.11g speeds. No fancy dual channel bonding, or 40MHz wide signals or "Turbo" modes. Just plain old 802.11g with some MIMO thrown in for better range. I find this results in the least amount of difficulties and provides the best compatibility while retaining a fair amount of speed still.

-Dan Spisak

And my Belkin "pre-n" continues to work very well indeed at Chaos Manor.


Thanks Jerry for so many years of great stuff.

AND MORE IMPORTANTLY...thanks for still turning it out. We love every post and column.

I don't know what it is...but your formula is wonderful and endlessly readable. You're our intelligent, tech-saavy "self"...forging into the "wilds" (often everyday wilds) of computing. We can't get enough.

All best and appreciation for good work.

Al Just a Reader Santa Monica

And thank you for the kind words


Subject: Camera envy

Dr. Pournelle,

I'm drooling. New Nikon D80, 10mp sensor and 2 brand new lenses that are within an enthusiast's price range. Price for the body alone is about $999, and with a brand new 18-135mm (about 28-200ish compared to 35mm film lenses) it is only $1299. A new 70-300 vibration reducing lense with their special lense coating will retail under $700, so you can get a complete 10mp setup from close-up/wide angle to 300mm VR zoom for around $2,000.

http://www.dpreview.com/news/0608/06080903nikond80withpreview.asp

This is going to put a serious dent in Canon's lineup, and it will be an awesome first digital camera for someone who cares enough about photography to have a 35mm SLR but wasn't sure how or when to transition to digital SLR. I'm not kicking myself for buying my Nikon D50 6 months ago since I specifically bought at the lowest end of the market so my initial investment would be minimal if I ever decided I HAD to upgrade, but if I didn't have a D50 already I'd be first in line for this setup. Nice rig for the non-pro enthusiast, no doubt about it. This would make a PERFECT airshow camera and lense system... For under 2 grand, you can get a great new camera, 2 lenses that cover your zoom from 18-300mm (25ish-400ish relative to 35mm lenses) plus vibration reduction in the long zoom lense, and 10mp is a great resolution for poster size prints.

Yea it's not your usual fare but especially with the death of the good tech shows, it's good enough news for ANYONE who uses SLR cameras that I thought I'd share.

A related orchid nomination for the guys who sold me my Nikon D50, 2 lenses, flash, filters, and assorted accessories - Cameta Camera is one of those "good" ebay retailers who really deliver what they promise. They have great *honest* prices, and since they are a real brick/mortar store that just happens to use ebay as their online storefront, you can call their 1-800 number and customize your order if you don't like the packages they put together on their ebay front. They did a great job with my order and they have a great positive feedback ratio from thousands of customers. I figure since you recently posted your frustration with an online camera scam operation, you might want a pointer to one of the honest camera retailers out there. Plus they beat the retail price on my order by over $400 without resorting to sneaky refurbished parts or scams, not bad IMHO.

Sean

Captain Ron Morse says "Pfui, but then I am committed to Canon lenses and DSLR bodies. In truth it's mostly a religious argument. Just ask any Hasselblad owner."

My Panasonic DMC FZ30 "Lumix" remains the main camera here, and is a great deal more than satisfactory, but I don't do really high end photography. On the other hand, George Margolin does, and I got the Panasonic on his recommendation. As I said in the column, I had some problems with some New York Internet Camera stores, but Abe's of Maine delivered what they promised with no difficulties at all.


To Be Continued...