Posts by Richard Plinston

Re: Meh

> Why rewrite? Emulate the hardware, and run the original software.

The question was about learning OS design and implementation. Merely running an existing OS doesn't do that. Rewriting an OS while solving different hardware problems does. Porting an OS for which you have source code (eg Minix) would also.

Re: Meh

> it would be rather nice if there was a small, cheap board available for the purpose of learning of exactly how Operating Systems work.

I am not sure why you think that is not already available ? Of course it would depend on which operating system(s) you would want to learn. MVS, or even OS360 may be difficult to rewrite for a Raspberry Pi, but Minix should work. It should be possible to write a p-code or q-code interpreter to run the Oberon OS on RPi.

Re: Author comment -- could you lot miss the point any more widely?

> You don't teach schoolkids woodwork with chainsaws and 100m tall trees.

You also don't teach woodwork by restricting the tools to be entirely made of wood.

Re: Author comment -- could you lot miss the point any more widely?

> Oberon shows that an essentially one-man language, compiler and OS can produce a real, viable, practical, usable OS that an entire University department ran on for decades.

C and Unix was originally "an essentially one-man language, compiler and OS". Minix was a one-man OS. Linux was initially one-man OS, so was CP/M. MS-DOS started (QDOS) as a one-man OS (though it was a copy of CP/M).

And what "University department" was that? Computer Science? Computers are not just for teaching about computers, they are tools for using to learn about many other things.

> The IT industry assumes that operating systems have to be written in C to work -- wrong -- and must by nature be big and complex -- wrong.

The IT industry has written operating systems in many different languages. Burroughs used an Algol derivative called ESPOL for MCP. ICL VME B was written in S3. CP/M was written in PL/M. The core of AmigaOS was TripOS which was written in BCPL. BCPL begat B which was the inspiration for C.

While OSes do not _need_ to be written in C, it happens that C produces fast running code that can be ported to many architectures. Other languages may provide different advantages, but the 'IT industry' is not there to make computers for university departments studying computing (they can do that themselves), it is there to make money by selling computers to do many different things, at different price points, and in different ways.

C achieves that.

> We should give kids something small, fast, simple, clean, efficient. Not piles of kludge layered on top of a late-1960s hack.

The Pi Zero is not locked into any particular OS. It already runs several: Linux, BSD, RiscOS, Win10IoT (on Pi2 only). There is no reason at all why others could not be ported to it.

> Using them, someone could port the OS, sure. It has not been done, partly because, as I said, it's an effectively obsolete set of tools that has been long superseded.

Fine, so what are you suggesting ? Or is it just that you hate Linux, C and Python but don't have any real alternatives ?

Re: Meh

> The article is not about programming languages. It is about operating systems.

Actually the article was about "tiny, inexpensive, single-board educational computers", though the OberonStation is not cheap. It mentions several boards, mostly Pi clones but also the BBC:Bit, CodeBug and C.H.I.P. As you say it did originally have the PyBoard.

You may _want_ it to be about operating systems, but it wasn't. The Pi (which more that half the acticle focussed on) can run many systems. It did discuss RISC OS (which is available) because that is the origin of the ARM processor.

> Thus, it doesn't matter what compiler is hosted on top of Linux, that does not address the issue that Linux is a problem, as it is too large, too complex, too slow, too cryptic and too difficult to be a good educational tool.

You appear to be obsessed about teaching 'Operating System Design and Implementation'. That is a rather niche area in University courses. Certainly there have been several operating systems developed within the constraint of being all done in a single language. But this is hardly a place to start teaching computing.

If you want to lock down a system so that it presents a simple, easy, programmable interface for teaching then that can be built on top of any operating system. The advantage of that is that it can be changed to run different teaching systems.

But these single board computers, such as the Pi, are not _just_ to teach computing in a theoretical way. They are a tool to learn _other_ subjects. They can teach electronics. In a physics lab they can be used with sensors and relays to learn about gravity, electricity, magnetism and much else. They can be used to learn about photography, and can do time lapse or high-speed for chemistry or biology, or even meteorology.

Re: Meh Python? Bias pascal?

> FAR better than BASIC, which was cut down ForTran intended to be a stepping stone to Fortran.

There was a language implementation called BASIC-FORTRAN which was a cut down FORTRAN, but this had nothing to do with K&K BASIC.

There are always some similarities in computer languages, arithmetic expressions generally are the same, unless it uses RPN.

Re: Meh Python? Bias pascal?

> Python, which assumes you already are an expert programmer.

I am not sure why you think that, perhaps you could explain.

Re: Meh Python

> the acronym is actually a backronym made up after the event.

Aren't they all ?

Re: Meh Python

> It's total rubbish for TEACHING programming.

You obviously have a huge bias towards Pascal and derivatives.

The essence of learning programming is fast feedback. This is why BASIC was designed as a teaching language (Beginners All-purpose Symbolic Instruction Code). Each part of the language could be tried out and results were immediate.

Pascal was also intended as a teaching language but it was not interactive it was compiled. The compiler was only to p-code so it was fast, but the whole cycle of edit-compile-run gave a slow feedback. Pascal also was very limited (deliberately) with no modules, limited IO and no data structures beyond arrays. After learning Pascal one had to change to a different language.

There were many extended Pascals, each incompatible with the others: UCSD, Turbo-Pascal, ISO Pascal, MS Pascal, ... and many derivatives, in fact as many variations as there were of BASIC. Both languages had brief popularity in the 80s but faded in the 90s.

Python has been around since the 90s and has had a single development path with few deviations. It can be used like BASIC as an interactive learning tool or can be used as a highly modular object-orientated language for very large projects - without relearning another language.

The main complaint about Python is that it uses indentation to represent the logic levels (rather than begin-end of brackets of some form). This has the advantage that the user and the compiler see the same logic levels - as long as the editor does not insert tab characters that will disrupt the indentation. Get a decent editors that recognises Python's needs (there are many) and there are fewer issues with understanding the program flow than other language's need for punctuation.

Also Python has just one 'style'. Other languages 'discuss' where the begin-end or {} should be placed, the variations in styles can obscure the logic from the beginner.

While Pascal may be useful for teaching how to program linked-lists or hashed arrays (because you need to program these yourself), computing has moved on from that and Python provides higher level data structures (lists, dictionaries, tuples) to solve _application_ problems rather than getting stuck in computer problems.

Re: Click-Bait much?

> A computer that's Pascal through-and-through would be awesome,

The ICL PERQ (Three Rivers hardware) of 1982 ran 2 operating systems: PNX which was Unix version 7; or PERQ which was written in Pascal and ran Pascal and Fortran. It ran 'Q-Code'.

Of course standard (Wirth) Pascal would be completely useless for writing an OS so there were several extensions to allow modular programming and communication.

> In that sense for most users it would be just as functional as their desktop PCs.

Well, maybe as functional as Windows RT was, but without the apps.

Re: So...

> Microsoft have been very good at providing good updates and upgrades over the years,

You mean like Windows Mobile 6.x -> Windows Phone 7. WP7 -> WP8. It seems that only _some_ WP8 models will get W10M, maybe fewer.

Re: Simplistic analysis

> MSFT aren't in the bsuiness of selling mobis, their strategy is to create a new product category.

I doubt that Microsoft have ever created a "new product category", they copy what others are already doing and sprinkle 'innovative' over the marketing blurb.

Re: Uh oh, El Reg...

> But it also runs the same familiar Windows (10) OS as a Lenovo or Asus laptop..

It may be that it has the same 'modern' UI, the one that Windows 8.x uses seem to dislike, but it does not run 'the same familiar' GUI that laptops run (ie the Windows 7 one).

> and can connect to a monitor, keyboard and mouse (wirelessly if necessary) to give you a desktop computer experience when you need it.

This only runs on two models so far - The 950 and 950XL. As these have ARM CPUs it is not really a 'desktop' experience, it is like RT but with fewer apps so far.

Microsoft confused users with Windows RT - Windows that could not run Windows programs. Now, by claimed that the 950 can give a 'desktop experience' they are continuing to perpetrate that same confusion. They also caused much confusion by announcing that Windows 10 would run on a Raspberry Pi. It just wasn't what most expected.

Ubuntu has had phones that can plug into monitor and keyboard/mouse and run desktop. It can run proper desktop applications, even on ARM.

Re: Does anyone think this is a good thing?

> One of the biggest killers of choice is "free" software. So Microsoft destroyed the browser market by bundling IE "free", with negative consequences still being felt thirty years later.

It was not that it was 'free', it was that it was bundled and compulsory. Spyglass wrote IE under a contract that paid them a few dollars for every copy sold. MS gave it away for 'free' and that destroyed Spyglass. One question is: does 'bundling' make it 'sold' as part of the product regardless of saying 'free' on the box?

When Win98 was installed the first reboot asked if you wanted to install IE but the 'Cancel' button was disabled. There was no escape, it had to be installed. Prior to IE, OEMs were installing Mozilla and Trumpet or similar. MS gave an additional discount (said to be $5) to _not_ install competitors.

So it wasn't 'free' that destroyed the browser market, it was 'compulsory' and 'OEM discounts'. MS also deliberately added non-standard features to FrontPage and IE so that Mozilla did not display 'properly'.

> Android with regard to updates and security is a result of the "free" nature of Android

There is a core of Android that is free, but if Google services are to be included then there is a charge. Google makes updates and new versions available to all, it is the makers that have to build and issue updates. Choose a vendor that does that.

> At the root of the problem of "free" are three simple issues: 1) that new entrants are unable to create any revenue stream to give them traction thus crushing variety, 2) the lack of successful new entrants reduces pressure on incumbents to innovate and maintain, and 3) the lack of revenue streams from user-choosers reduces the incentive to keep the dominant "free" products secure and up to date.

That is a very shallow analysis. Red Hat (as CentOS or several others) is free, Ubuntu is free, as are many others, yet they have entered the market and created a revenue stream, _and_ they keep the products secure and up to date.

Re: Does it infect Linux destops?

> on average take longer to get patches released once publically known (more days at risk).

That is because Microsoft does not make public the issue until the patches are released. This does not change the 'days at risk' as you claim because it is not the 'public' that create the risk. The risk occurs from the time that black hats create it, not when the public is told of it.

Re: Does it infect Linux destops?

> Downloading a random executables from an untrusted (not to mention non-ssl) source just isn't done.

Not only that but on Linux a downloaded file is _not_ executable no matter what its filename is, it requires specific user action (eg running chmod) before it can become executable. Also, Linux mail program do not automatically execute attachments, which is another vector on another OS.

Re: 5 years is a long time

> What predictions for 2015 were this soothsayers making 5 years ago?

http://www.pcworld.com/article/230151/idc_windows_phones_to_overtake_iphone_ios_by_2015.html

Re: first tablet?

> 2002 (and SP in 2005)

Actually a decade earlier:

http://haverzine.com/2013/12/20/microsoft-mobile-part-1-the-birth-of-pen-windows/

Re: Overpriced

> Or just remote in.

> Huh !!!! there is a link missing in that chain that he is complaining about.....

You can SSH in using the USB connection, or use a USB 'Terminal Cable' to GPIO pins.

http://elinux.org/How_to_use_an_Android_tablet_as_a_Raspberry_Pi_console_terminal_and_internet_router

https://learn.adafruit.com/adafruits-raspberry-pi-lesson-5-using-a-console-cable

Re: Overpriced

> a compute with _NO_ network connectivity is rather useless in this day and age.

And yet many people find Arduinos to be useful.

This is a 'maker' device. Build it into a robot, wear it on your coat, run your plant watering system, ... Development can be done on-board (with TV and keyboard) or by connecting using SSH. Or develop on Pi 2 and deploy to Zero.

Think of it competing with the BBC thing or Arduino rather than a PC - it is only $5.

As an IoT device it can use 1-wire or I2c. You could add Bluetooth or WiFi.

Re: @fung0

> I do agree that the Surface is a good laptop.

I disagree that it is a good _lap_top. It may be fine when used on a flat surface, such as a desk.

For putting on a lap: the weight distribution is wrong, the length between front of keyboard and back of stand is too great, the screen size is too small for when it is positioned at the knees, the screen angle has inadequate adjustment, the keyboard/screen joint is too flexible, the edge of the stand digs into the flesh if shorts are worn, it is too unstable to use for touch, ...

Re: Very interesting article.

> The obvious solution for Microsoft would have been to evolve Windows CE/Windows Mobile, and adding a touch-friendly UI.

Windows Phone 7 was CE based with a touch friendly UI. Look how well that worked.

CE was the MS-DOS of the ARM world. It could not support more than a single core (which is why MS asked such inane questions as "why would you need dual core?"**. It had no multitasking capability (hence 'tombstoning'), it had something similar to MS-DOS TSRs.

** Dual, Quad and Octocore is not so much about computing power as battery saving - cores can be turned off, the more you have the less battery draining standby is. This is why MS 950XLs have Octocore.

Re: The strategy of stupid

> How is City of Munich going with their migration back to Windows/Office?

It never started.

Re: Very interesting article.

> these devices would have been impossible to produce until relatively recent years never mind what OS you might install and whether or not it was fit for the purpose.

The Nokia/Maemo tablets were available to buy 10 years ago this month. I still use my N800 occasionally.

Re: "said the biggest impact Uber was having is traffic congestion"

Introducing a forced 5 minute delay will increase the average trip time (though by less than 5 minutes). To achieve the same number of trips per day will thus require _more_ cars. This will increase the congestion.

In theory the Uber mechanism should decrease wait time, and thus average trip time (when calculated from customer starting to want to make the trip) compared to other mechanisms. Thus there should be a decrease in the number of cars required for a given number of trips. However, having a more efficient system increases the demand. For example, if the service time drops below the time taken to walk to the destination then more will take the service.

Re: "While arguably less efficient than a command line"

> CLI too in many cases can be very inefficient.

Microsoft went out of its way to make the CLI useless. While DRI gave CP/M-86 through DR-DOS (and all its other OSes) decent command line editing and recall, MS did its best to make its inefficient. Windows 95/98 did actually have doskey.com that gave a halfway decent command editor but it was not installed by default and not even in the manual.

At the 1983 COMDEC, Paul Allen outlined what features the 'next' version of MS-DOS (2) would have, including a help system (finally in MS-DOS 5) and command line editing. But they saw GEM being demonstrated and then started writing Windows 1, dropping most of what Paul announced.

> the DRDOS/Win3 incompatibility was the old-MS at its worst -

What makes you think that these tricks are only done by 'old-MS'. 'new-MS' just hide them better.

> refused to allow DR (as a rival) to have pre-release access to win3 and then used an undocumented response from some DOS command to cause it to crash on DRDOS

Actually DRI did have pre-release access, and had it running fine on DR-DOS, but when Win3.x was released MS had added the AARD code. This produced warning messages that it wouldn't work correctly. DRI rushed out an update that defeated AARD, but, even though Win3 ran better on DR-DOS, and had much more memory available, the damage had been done.

Re: Am I splitting hairs

> Windows 1.0 (all the way up to 3.1.1) wasn't an operating system but just a DOS GUI?

Windows 1 was more of a graphics library with a few simple plug-ins. It could be static linked into your application so that it could run on a bare MS-DOS system. This was used by PageMaker:

"""Until May 1987, the initial Windows release was bundled with a full version of Windows 1.0.3; after that date, a "Windows-runtime" with no task-switching capabilities was included.[10][11] Thus, users who did not have Windows could run the application from MS-DOS."""

Re: Ahhh Windows 3.11 for Workgroups...

> at the time we called this 'Widows for Wombats'.

Later 3.11 became 'Windows for Warehouses'. The new Windows (later called 95) had been announced and was supposed to be delivered by April but was delayed. Buying of 3.11 all but stopped though it was still being manufactured*. It was reputed to have filled several warehouses before manufacturing eventually** changed to Win95.

* Previously, when OS/2 Warp was about to go into full production, Microsoft went to all the diskette manufacturers and bought the total production for the next six months. This restricted the availability of OS/2 and was still being used when Win95 went RTM.

** Bill Gates is reputed to have said that "Windows 95 will be out before Christmas, but we may have to delay December for a couple of months".

Re: "While arguably less efficient than a command line"

> Novell DOS 7 had pre-emptive multitasking

DRI had pre-emptive multi-tasking, multiuser and networking systems since 1978 with MP/M, MP/M2, MP/M-86, Concurrent-CP/M-86, Multiuser-DOS and various other derivitives. It was demonstrating Concurrent-CP/M-86, with full pre-emptive multi-tasking using virtual screens*, when MS announced MS-DOS 2.0.

DR-DOS was originally built from Concurrent-DOS source code by removing multi-tasking and multi-user. It was preceeded by DOS+ which left in some multi-tasking.

DR-DOS 6 added task-switching so multiple tasks could be loaded and switched between using EEMS (only the foreground task would run). Novell-DOS added back multi-tasking but it was a very poor imitation of Multiuser-DOS.

> It basically could do most of what Win 95 could do, except that it was command-line and not a GUI,

DRI's GEM GUI had sold a million copies before Windows 1 was released. It then wound up on Atari 512s as TOS.

* A keystroke combination switched the screen between the running programs. On an IBM PC 8088 it required an EEMS memory card, such as a AST RAMPage. To do the context switching. It only required a small register block to be changed to 'bank switch' the program. The OS could recover disk access time to give CPU time to other programs (which Windows 3.x couldn't do).

Re: No comprende

> You can copy the SD cards without needing a windows PC,

Not just with W10IoT on a RPi2.

> and you can install / uninstall / change boot app via a web browser running on any other computer.

By "change the app" I was indicating an edit/compile cycle. But even changing the boot sequence to run a different app requires another computer.

> Microsoft has never suggested that W10IoT will be a drop in desktop OS.

When Microsoft announced Win10 for RPi the news boards were full of:

"""Microsoft says it’s "delivering a version of Windows 10 that supports Raspberry Pi 2. ... With the pricing of the Raspberry Pi 2 and Microsoft’s free copy of Windows 10, you could have a full PC for just $35 later this year. We’ll have to wait to hear more information from Microsoft on how Windows 10 will function on the Raspberry Pi 2, but the company says it’s planning to reveal more "in the coming months." It's likely that this version of Windows 10 will only run modern universal apps, as the Raspberry Pi 2 includes an ARM-based processor."""

Now, I knew what MS had done with the Galileo board and that RPi would be much the same, but MS did not make this clear, nor did they with RT. Many believed they would be running Photoshop on their RPi.

Re: Raspberry Pi 2B is not an IoT device

> the Pi originally had 2 core 800Mhz and 512Mb.

No. The Pi originally had a single core 700MHz and 256Kb.

"""performance is similar to a 300 MHz Pentium II of 1997-1999"""

This is more than most used for Windows 95 or 98, or OS/2.

Re: No comprende

>> The Windows IoT thing doesn't even have a display.

> That's strange, I've got mine hooked up to a screen and its quite happily rendering to it.

Windows 10 IoT does not "have a display". UWP _apps_ running on W10IoT may "have a display".

That is: there is no OS GUI.

> I wish people would criticise MS for the stuff they do (there's enough of it) rather than making things up. Like the guy down the page who says you need a windows PC to drive the Pi when you use W10 IoT - you don't. You only need one for the initial imaging. After that they run stand alone.

A full Windows 10 PC is stated as a requirement. This is required _each_time_ you image the RPi2 SD card and each time that you need to change the app.

As a comparison a RPi2 with, say, Raspian can be used to develop the app as well as run it, no 'PC' required at all, and can even be used to develop Arduino apps and load then to the Arduino.