Dr. Jerry Pournelle

Email Me

Why not subscribe now?

Chaos Manor Subscribe Now

Useful Link(s)...


Hosting by

Powered by Apache

Computing At Chaos Manor:
The Mailbag

Mailbag for April 23, 2007
Jerry Pournelle jerryp@jerrypournelle.com
Copyright 2007 Jerry E. Pournelle, Ph.D.

April 23, 2007

Continuing from last week on readers for the blind;

Subject: Blind Services

"what is the impediment, while we wait for the machines to improve, to having highly-educated third-worlders (or even, gasp, American college students) read the NYT and news, op-ed, etc each morning, to be RSS'ed to anyone, sighted or non-sighted, who might choose to subscribe?"

It has been done for decades over the FM band, though you will need a special radio to receive it. They read the daily newspaper and other publications. Allegedly some stations offer "podcasts" of their daily programs.

One other product that I forgot about till just now is a wearable monitor (they look like thick sunglasses) with a portable zoom camera. The camera in wide angle mode has a low enough focal length to allow one to navigate the sidewalks with the zoom feature to aid reading signs, especially bus numbers. The monitor is low resolution for someone who has no vision problems but more than fine enough for those with macular degeneration.

Gene Horr

I am not surprised that all these services exist, but I didn't know about them. A quick Google on the subject reveals a bunch of them including WRBH, this PR Newswire item, and even a Wikipedia entry.

Subject: OCR artifacts really a bane?

Hi Jerry,

Nothing to add to what Mr. Early said. That is how I get my best results too. That said... I'd like to throw an idea at you and the gang.

With the exception of a few exceptional documents, like the reader or your novels, does absolute accuracy really matter so long as the captured text is tied in some manner to the scanned image so you still have it to refer to? (Like Acrobat or DevonThink does for example in the PDFs they generate.)

I have scanned literally hundreds (strike that, thousands if you count various conference proceedings CDs I put in too) of papers in to DevonThink over the last few months in order to get my professional library into some format where I can actually find something or to use the AI features to find linkages I haven't noticed. The bulk of these papers also went on the office LAN as PDFs. With either DevonThink or one of the desktop search engines (Google, Windows, Spotlight et al) a few key words carefully chosen by one skilled in the art allows one to drill down to the pertinent literature in a very short time indeed. This is especially true in comparison to my prior method of following my vague memory on a bug hunt through the file cabinets. For this purpose I really don't care if the OCR phase leaves artifacts. The keyword search has enough candidates, in my experience to date, to get me where I need to be. For example I was able to find two papers written in the late 50s that gave a customer the assurance needed to proceed with processing a million bucks worth of parts that they would have otherwise scrapped.

All the best,

Richard Kullberg

I can see how documents linked to a bit map image of the original could be useful for many purposes, but I would hate to have someone reading one of my novels, or an essay, and have to connect on-line to make sense of it. I suspect that would not do a lot for ebook sales.

But I will agree that when it comes to finding information in a sea of data, anything is better than nothing, and your key word searches are a good example of that.

Subject: Regarding Registry Cleaning....

Don't even get me started on this one. I have NEVER had any luck with the myriad of tools out there that claim to make things better by cleaning up the registry. They have always caused me more grief than I had before I ran the tool. Plus I only ran the tool because I likely had a problem in the first place.

This goes back to my biggest gripe of all. Microsoft needs to finally admit that the entire registry concept was a big pile of pyrite from the start. It's time for this failed design to be put out to pasture and for us to get back to the greater stability and easier to control OS we used to have (though now it's becoming a distant memory).

The registry could even be phased out in stages. My first stage would be to go to a fully-dynamic registry. Create a "Registry" folder and fill it with .reg files put there by whatever application has made the mistake of thinking that the registry is a good, sound way to go about storing variables. The OS could boot, process the registry folder and thus create a fresh registry with every boot. In fact, only Microsoft should be able to store .reg files in the root Registry folder. Everyone else must put theirs in a subfolder under it. I'd even go so far as to say that even Microsoft shouldn't put any others in there except for the core OS but that's just wishing for the impossible.

An application should only be able to write to its OWN .reg files. It should be considered bad form to edit a registry file from another application. So, if I go and install some new Intuit program, it can go ahead and create 35 .reg files if it so desires. It can write to them all it wants. It can screw them up all it wants if that's what it wants to do for that version. I then have the ability to uninstall the application and all of its .reg files and, presto, my system is running smoothly again. At the current rate this will be a story I'll need to tell my future grandchildren and they'll laugh like my son does now when my father tells him about the man who used to go door-to-door offering waffles for sale.

Rich Heimlich

The Registry as Watkins Man... You do know that the last Watkins Man has a web site...

I quite agree that the Registry needs adjusting, probably with an axe.

Returning to a previous subject

Subject: Read aging of flash memory

Jerry: When you read a file from flash memory with linux, it will update the access time, which qualifies as a write. This behavior can be eliminated by mounting the flash drive with the noatime switch. I don't know how access times are logged in Windows, but if I do a directory listing (using linux) of our SAN (MS Server 2k IIRC) using the -u switch (sorts by access time) it changes the order of the listing, and if I access a file at the tail end of the listing, and re-list, it moves to the head. I would expect access times to be written to the file header by the OSX operating system.

Christensen, Chris (Aspen Research)

PS: MS explanation of NTFS logging of access times.

"Each file and folder on an NTFS volume contains an attribute called Last Access Time. This attribute defines when the file or folder was last accessed, such as when a user lists folders, adds files to a folder, reads a file, or makes changes to a file. The most up-to-date Last Access Time is stored in memory and is eventually written to the disk in two different locations. One is within the file's attribute, which is part of its MFT record. The second is in the index of the directory that contains the file. "

Christensen, Chris (Aspen Research)

Which means that reading a file from a silicon disk will actually write to that disk. I doubt that makes any difference in the life of a silicon disk; almost certainly not enough to cause it to wear out before it's obsolete.

Subject: regarding "the mythical man-month"


There is what I call "the scariest fact in engineering."

If you assume (a) a 40-hour productive work-week, and (b) that every person on the product team is required to spend six minutes (1/80 of manday) reviewing each man-day of productive output to assure project consistency, then the MAXIMUM productivity of the team is 400 man-days of productive work per week. And the point of diminishing returns arrives quickly.

The operative equation is

M = NH / (W + (N-1)R)


M = productivity in work-days
N = Number of Employees
H = Number of hours worked per employee per week
W = hours per work-day (8)
R = Review time per work unit in hours (0.1)

Some observations:

With 21 employees, each employee spends 80% of her time in productive work and 20% of her time reviewing other person's work (think meetings).

With 81 employees, each employee spends 50% of her time in productive work and 50% of her time reviewing other person's work (think meetings about meetings).

When you reach 321 employees, each employee spends 20% of her time in productive work and 80% reviewing other employees work. It is necessary to add five new employees just to produce one man-day additional productive product.

This does not promote .. anything.

The job of GOOD management is to improve on these figures instead of succumbing to the committee syndrome and making them worse.

(And I won't tell you what I'm doing now -- other than waiting for my tax accountant who is running two hours late; it's too depressing.)


When I was in the aerospace industry during the Cold War, the 40 hour week was mythical. I recall one evening when a blivet came in to Aerospace Corporation from the Pentagon on a Friday afternoon. I told Mr. Dorrance, my director, "They want it Monday, but Monday morning we have a meeting with the General. Sunday afternoon Stevenson and I are getting the rfp out. And if I am not home Saturday my wife will leave me." Of course the answer was, "So what are you doing tonight..."

Adding new people to our Project 75 Team was very difficult, and once we were well underway became impossible. Mythical man month indeed.

Von Karman used to tell us "Is taking one man and one woman and nine months to make a baby, and putting nine more men on the job does not reduce that time."

Subject: iPods, lossless, battery life, etc..

Dr. P:

Last week's mailbag and its discussions of flash & iPods was quite interesting... Some observations of my own:

My "ancient" 2-year old 20GB iPod (with hard drive) seemed to naturally migrate to the car as soon as stereo equipment was available to control it "naturally" via the connector in the bottom. And then, a year later the stereo (but not iPod) was stolen and now using my latest Alpine is as fast and natural as using the iPod interface itself. The iPod stays charged and on those occasions when I fly I just pull it from the car and have portable music. But more than 99% of the time it's just the way I carry a copy of a subset of my music collection around. I have a flash-based iPod for the gym & hiking that weights nothing and is essentially indestructible.

Getting to my point - in the car and in they gym I use MP3 format for the music, as there is enough external noise to make higher fidelity rips pointless. But even with MP3s, if one uses "variable bit rate" ripping, the quality goes up enough that my hearing can detect it.

At home I use Apple Lossless as my format. This is a truly zero-loss compression and, usefully, it captures all the meta-information for your music in each file... allowing you to lose your iTunes .xml file and still have all the information in the music files.

With a little bit of clever iTunes work, one can rip ones CDs to a tree of files in lossless format, change the destination directory and convert the lossless tree to VBR MP3s. A warning about this, though - the 5000+ songs I had in lossless format took 24+ hours to convert to MP3s on my old 2.4Ghz P4.

Rob Philip Boulder, CO

While I would feel deprived were I confined to a chair for more than a few minutes without something to read, I have never felt deprived if I have nothing to listen to. I did for a few months subscribe to a taped business news service, but I found I almost never listened to the tapes. I own several models of iPod, and at one time I had the Creative Labs competitive item, but I gave most of those away for lack of use.

When Niven and I hike on our fire trails in the California city wilderness parks — there's 50 square miles of park about a block from my house — we often see people oblivious to their environment. They have iBuds in their ears, and they never respond when we wish them a good day. Sometimes we curse them, but they don't hear that either. But we talk to each other, and listen to the birds.

I have listened to murder mysteries on tape when I take long trips, and I'll generally have KUSC playing on the car radio when I'm alone, but I can survive without entertainment. It's probably a defect; certainly I seem to be part of a vanishing breed. What I find I can't survive without is my Olympus WS-100 pocket recorder for taking notes when my hands are not free.

Continuing a discussion of operating systems and compilers:

Subject: Well not really accurate.

You said "and no operating system should be released if it has been compiled with C or other compilers that will compile nonsense."

Linux, and OpenBSD are both written in c. They both have a good security record compared to Windows. One of the most secure OS's I have ever seen is VMS and I think that was written in assembly. Linux and OpenBSD are not perfect but they do tend to do better than Windows. The real trick is not to write nonsense. Windows is full of nonsense. A good example is that you almost have to run with administrator rights as a Windows user. Under Linux and other Unix like OSs you run as a user. The other bit of nonsense is that it is over integrated. Everything that Microsoft writes uses deep hooks into the OS. Since when should a web browser be vital part of the OS?

While I am all for range checking I just don't think you can write a usable OS that way. Every now and then you will have to get down and bang the bits. How for 99% of the applications and utilities that you run on the OS then yes "safe" development environments are the way to go.


I replied "There should be no buffer overflow exploits in an operating system. None." That drew this response:

You are correct. As far as I know VMS hasn't had any. The problem comes to what is part of the OS and what isn't. Buffer overflows in the kernel are very rare even in Windows. The latest stack overflow in Windows happened not in the kernel but in a library that handles animated cursors! Microsoft's Operating System tends to magnify the problem because:

1. Most people are running with full admin rights.

2. A large number of people running Windows Machines know nothing about computers and don't want to know anything about computers.

3. There are so many of them they make the biggest target. Think of the Irish Potato famine, Microsoft actively discourages any type of technical diversity. What percentage of computers today run Outlook, IE, Office, and Windows? How many have at least one Visual Basic program installed?

As I said for things outside the kernel I agree safe development environments make sense. In the kernel they may be way too restrictive. It is possible to write safe code in c. The problem is way too many programmers make no effort or ignore the compiler warnings.

I will say that if you want the most secure OS that has a good number of applications you better format every machine that runs an OS from Microsoft and start loading OpenBSD. While it doesn't have a perfect security record it is one of the best and you have a usable selection of software.

I agree with you but I doubt that even the best of tools will keep a lazy programmer safe forever.

To which I can only say I am not convinced. Even the best programmers have moments of inattention. Computers are very good at doing what they are told to do. That means that if they are told to do range checking and strong type checking, they do that. They do it even when the programmer had a bad day, or decided he was clever enough to get away with bad practices, or simply took leave of his senses.

Niklaus Wirth tried to design compilers that were fussy. It was sometimes difficult to get programs to compile but once they did, the programs didn't need a lot of debugging: the program generally did what you expected it to.

I understand that strongly structured programming languages do not produce code as "efficient" as C. On the other hand, C compiles nonsense and allows all kinds of exploits.

The Pentagon tried the ADA experiment. That went wrong for reasons that would take to long to give here; but the idea that all military software ought to be written in a strongly structured language was a good idea; and it's a good idea for operating systems as well. Once the program is written and compiled it's possible to go in and hand optimize frequently used loops to increase efficiency.

Subject: Feedback on 8 core Mac

I was reading about the 8 core Macs on your site and telling my partner about them. She asked me what good it would do her. I explained that with the right program it could cut down the time of her DVD authoring program from a couple of hours to possibly a few minutes. Her reply was what good is a hobby if it doesn't take any time?

Kerry Brown

And to that there is no answer. Thanks!