merlyn@stonehenge.com (Randal L. Schwartz) wrote:
"Joshua" == Joshua 'Schwa' Gargus schwa@cc.gatech.edu writes:
Joshua> Apple was one of the original licensees of Smalltalk-80 when Joshua> it was first release outside of Xerox PARC.
As was Tektronix. I first used Smalltalk on a Magnolia (68K) at Tek in 1982.
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
-- Hannes
Hannes Hirzel hannes.hirzel.squeaklist@bluewin.ch wrote:
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
Almost none and very few. The Magnolia was a predecessor of the commercial 4404 which had a massive 1Mb of ram. And a stunningly fast 8MHz 68000 cpu.
So, why was it that we couldn't do a fullscale Smalltalk for a palm pilot? Errrrr.....
tim
On Mon, 17 Mar 2003 10:30:34 -0800, Tim Rowledge tim@sumeru.stanford.edu wrote:
So, why was it that we couldn't do a fullscale Smalltalk for a palm pilot? Errrrr.....
I think that has more to do with the OS architecture than the processor capability...
Later, Jon
-------------------------------------------------------------- Jon Hylands Jon@huv.com http://www.huv.com/jon
Project: Micro Seeker (Micro Autonomous Underwater Vehicle) http://www.huv.com
At 10:30 AM 3/17/2003 -0800, Tim Rowledge wrote:
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
Almost none and very few. The Magnolia was a predecessor of the commercial 4404 which had a massive 1Mb of ram. And a stunningly fast 8MHz 68000 cpu.
Actually, the 4404 (which was announced in July 1984 and shipped in Jan 85) had a 10Mhz 68010 (it had a virtual memory OS which plan 68000s did not support). It was available in 1 and 2 MB configurations. Smalltalk ran ok in 1 MB but for "large" projects 2MB was preferable.
Tektronix Smalltalk was based upon Smalltalk-80 Version 2 and my recollection was that the initial Tektronix release had about 250 classes.
I presume that Squeak was based upon Smalltalk-80 Version 1 as that was what was originally licensed to Apple (along with Tek, Dec, and HP). As far as I could tell, the primary reason for ST-80 V2 was the Xerox realized that the V1 license terms were, from their perspective, way too favorable to the licensees. V2 provided an opportunity for them to revise the license. There were a number of times at Tektronix that we regretted having updated to the V2 image and license.
Allen Wirfs-Brock
Hey Squeakers,
At 10:30 AM 3/17/2003 -0800, Tim Rowledge wrote:
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
And Allen Wirfs-Brock replied:
Almost none and very few. The Magnolia was a predecessor of the commercial 4404 which had a massive 1Mb of ram. And a stunningly fast 8MHz 68000 cpu.
While reflecting on the early days of Smalltalk, let's not forget the contribution of the San Diego-based Levco company -- Duane 'Squeaky' Maxwell being among the intrepid band of engineers who worked magic showing Apple the way. The Levco Prodigy 4 was a ground-breaking, price-busting machine on many levels.
Here is a page with pictures of our original Fat Mac and SE-based Prodigies:
http://squirrelfeeders.com/comic_strip_factory.html
A 16 MHz 68020, with a math coprocessor, 4 MB of contiguous RAM (man, we thought we were flying when the most memory we had in our Fat Mac prior to the Prodigy was 512K!)... and the first commercial SCSI hard drive interface (our first Prodigy had a whooping 10 MB INTERNAL hard drive... that's right, the Levco guys figured out how to safely and effectively fit the Prodigy daughterboard AND a 'chimney-based' internal hard drive into the original Mac case!)... This was truly an amazing and groundbreaking machine.
The capper was when you connected an eMachines (I think this was the first big screen to work with a Prodigy) large-monitor to a Prodigy. For all intents and purposes, this created a Mac-based workstation that was an order of magnitude cheaper than similarly powerful workstations at the time. (The original Prodigy boards were $10,000! Expensive by today's standards, but super cheap when you figure what they were comparable to when compared to scientific and engineering workstations.)
A minor tweak of the Apple Smalltalk display classes was all that was needed to get Smalltalk SCREAMING on a Prodigy.
I clearly remember the day that Timlynn and I made an appointment to visit the Xerox Pasadena research lab (the 'southern cousin' of PARC) and we had a chance to show them a Prodigy in action. Until that time, Xerox thought Macs were 'silly toys'. We put the Prodigy under a draped table with the eMachines monitor, keyboard and mouse on top. We ran Xerox's HUMBLE expert system and some Smalltalk benchmark programs and BLEW THEIR SOX OFF! The Prodigy was faster than their Dorado and whatever their commercial Smalltalk workstations at the time were.
When we lifted the drape and showed them the Mac below, they nearly fainted! :-)
Needless to say, Xerox's opinion of Macs changed after this experience.
Somewhere in my (physical) photo archive, I have a picture of the big screen Levco Prodigy 4 running Apple Smalltalk. If I come across it, I'll add it to the page above.
Smalltalk on a Levco Prodigy was awesome and is a pleasant memory whenever recalled.
Best Regards, Jim on behalf of --Timlynn and Jim-- Founders and Research Directors Sohodojo - 'War College' of the Small Is Good Business Revolution http://sohodojo.com Director and Entrepreneur/Futurist In Residence NARFI - North American Rural Futures Institute http://narfi.org
Jim Salmons and Timlynn Babitsky ______________________________________________ JFS Consulting, our Portfolio Life nanocorp http://jfsconsulting.com
While reflecting on the early days of Smalltalk, let's not forget the contribution of the San Diego-based Levco company -- Duane 'Squeaky' Maxwell being among the intrepid band of engineers who worked magic showing Apple the way. The Levco Prodigy 4 was a ground-breaking, price-busting machine on many levels.
Just for the record, nobody calls me "Squeaky". Well, except maybe Jim, just now. I understand the US Secret Service gets very nervous about people with the name "Squeaky", particularly the old-timers.
I moved recently, and I ran across the little stack of floppies for Apple Smalltalk-80, though I think I may have been missing one. I think I may even have a Mac lying around that can still read 800K floppies - I may even have a Mac SE in the garage that might run it.
Hmmm....
-- Duane
PS. I don't think we ever successfully sold a Prodigy upgrade for $10K - but there were a lot sold at $5K.
Duane Maxwell wrote:
I moved recently, and I ran across the little stack of floppies for Apple Smalltalk-80, though I think I may have been missing one. I think I may even have a Mac lying around that can still read 800K floppies - I may even have a Mac SE in the garage that might run it.
Hi Duane/Everyone,
You might know all of this, but if you're looking for a good way to archive 800k floppies you come across (as the old timers start rooting through their closets), there are two (similar) methods I can suggest.
First, before you start, find a machine you know can read/write 800k floppies (actually, any Mac with an internal floppy drive should still be able to read 800k floppies, but not write them after MacOS 7.6 or so). It'd be a shame to scrape a disk at this point. Also, write-protect the disks so a newer MacOS won't rewrite the desktop folder on each one.
Then, use either SunTAR or Apple's Disk Copy (the newer version should still run on System 7.0 if needed) and make disk images of the floppies.
Apple's Disk Copy is nice because it's easy to restore the images to floppy on just about any Mac, and you can also mount the raw images to attempt installation. If you choose this route, I suggest burning the images and utility to CD, so you can lend/save/etc. the images for future use while retaining the ability to install directly from CD and reproduce the floppies.
With SunTAR, if you read straight from the floppy device to a file you'll end up with a more standard disk image, like you would have if you could have used dd on *nix on a box that could have read the 800k disks (like a Mac running A/UX, NetBSD or Linux). This can be handy because the resultant files are flat, with no resource fork, so they can be mailed or archived easily, mounted using the HFS filesystem on Linux or NetBSD (and probably OSX, though Disk Copy images would work there, too), and possibly run under a Mac emulator (ARDI's Executor or BasiliskII maybe).
Either way, the best thing you can do is get the data off the floppies as soon as possible. Then, if your version of MacOS needs to rebuild the desktop on each volume (if they're System 6 disks), you can freely butcher a new floppy copy or writable Disk Copy image without worrying about any potential side-effects.
Good luck!
-Jack
At 8:33 AM -0800 3/18/03, Jack Johnson wrote:
You might know all of this, but if you're looking for a good way to archive 800k floppies you come across (as the old timers start rooting through their closets), there are two (similar) methods I can suggest.
I recently decommissioned my last Mac with an internal floppy drive. Before I did so, I read all of the floppy disks that I could find in the house onto my hard drive.
OS X does not support the internal floppy drive at all. It supports USB floppy drives, but these drives do not have auto eject, and do not read 800k disks.
The best program that I found, under OS9, is "Shrinkwrap". You can set Shrinkwrap into batch mode, where it will read a disk, create an image file, eject the disk and ask for a new one, all without keyboard or mouse intervention. This program is what made it feasible to copy all of those disks without driving myself crazy. Shrinkwrap does not seem able to mount the resulting images any more under OS9, but Apple's disk copy will do so.
On lunes, marz 17, 2003, at 23:03 America/Quito, Duane Maxwell wrote:
While reflecting on the early days of Smalltalk, let's not forget the contribution of the San Diego-based Levco company -- Duane 'Squeaky' Maxwell being among the intrepid band of engineers who worked magic showing Apple the way. The Levco Prodigy 4 was a ground-breaking, price-busting machine on many levels.
Just for the record, nobody calls me "Squeaky". Well, except maybe Jim, just now. I understand the US Secret Service gets very nervous about people with the name "Squeaky", particularly the old-timers.
I moved recently, and I ran across the little stack of floppies for Apple Smalltalk-80, though I think I may have been missing one. I think I may even have a Mac lying around that can still read 800K floppies - I may even have a Mac SE in the garage that might run it.
Hmmm....
-- Duane
PS. I don't think we ever successfully sold a Prodigy upgrade for $10K - but there were a lot sold at $5K.
Using Disk Copy you can read 400K floppies in MacOS 9.2 and create .img files, you can not mount the images in MacOS 8/9/X because the MFS file system is not supported in this OSs. I recently wrote a small Python script (my first program) to access this images, and list the directory, view the text files and copy (with the resource fork and fInfo), works with the version of Python installed in MacOS X 10.2.4. Is in the Public folder of my iDisk with the name 'utilMFS.py' (12KB).
Javier Diaz-Reinoso Web: http://homepage.mac.com/javier_diaz_r/
"Sohodojo Jim" salmons@sohodojo.com wrote...
A 16 MHz 68020, with a math coprocessor, 4 MB of contiguous RAM (man, we thought we were flying when the most memory we had in our Fat Mac prior to the Prodigy was 512K!)... and the first commercial SCSI hard drive interface (our first Prodigy had a whooping 10 MB INTERNAL hard drive... that's right, the Levco guys figured out how to safely and effectively fit the Prodigy daughterboard AND a 'chimney-based' internal hard drive into the original Mac case!)... This was truly an amazing and groundbreaking machine.
Yes, there was definitely something magical about this level of performance. We had the same experience inside Apple when we got hold of a couple of prototype "Big Macs" built by Rich Page that coupled a 16Mhz 68020 to a 1024x768 (monochrome) display. Until this time, much of our Smalltalk and UI work was guided by a sort of intuition about what should be simple and what not, but always we were sort of drumming our fingers waiting for the processor to complete these wonderfully streamlined operations (things like find all senders, add an instance variable, etc), and unable to really use more than a couple of windows on screens smaller than VGA.
Then, suddenly, all the design for interaction paid off almost overnight,with the arrival of that next generation of hardware. It was heady stuff. I remember Scott Wallace was doing a lot of the heavy work on Fabrik in our group at the time. He would typically work at home for a day or two and then come in to compare notes and exchange a demo or two. When he got his Big Mac, he simply didn't come in or call in for two weeks. Finally I called his home just to check that he was OK, and found him nearly expired. As I recall he said something like, although we had built this incredible system in which you could literally change anything and keep running, until that time processor speed and screen real estate had kept our ambitions in check. Now, at a single stroke, those barriers had simply vanished, and we could just go on doing one thing after another with nothing to hold us back except, ultimately, our ability to stay awake.
- Dan
Hi all
Thanks for all these notes. While reading them, I was thinking that they could make a really fascinating book. Dealer of lightning is good but a book really focused on the Smalltalk pionners would be really interesting. May be collecting all the material into a dedicated wiki would be a first step.
Stef
I know that inventing the future is more fun that chatting about the past but still...
On Tuesday, March 18, 2003, at 08:40 AM, Dan Ingalls wrote:
"Sohodojo Jim" salmons@sohodojo.com wrote...
A 16 MHz 68020, with a math coprocessor, 4 MB of contiguous RAM (man, we thought we were flying when the most memory we had in our Fat Mac prior to the Prodigy was 512K!)... and the first commercial SCSI hard drive interface (our first Prodigy had a whooping 10 MB INTERNAL hard drive... that's right, the Levco guys figured out how to safely and effectively fit the Prodigy daughterboard AND a 'chimney-based' internal hard drive into the original Mac case!)... This was truly an amazing and groundbreaking machine.
Yes, there was definitely something magical about this level of performance. We had the same experience inside Apple when we got hold of a couple of prototype "Big Macs" built by Rich Page that coupled a 16Mhz 68020 to a 1024x768 (monochrome) display. Until this time, much of our Smalltalk and UI work was guided by a sort of intuition about what should be simple and what not, but always we were sort of drumming our fingers waiting for the processor to complete these wonderfully streamlined operations (things like find all senders, add an instance variable, etc), and unable to really use more than a couple of windows on screens smaller than VGA.
Then, suddenly, all the design for interaction paid off almost overnight,with the arrival of that next generation of hardware. It was heady stuff. I remember Scott Wallace was doing a lot of the heavy work on Fabrik in our group at the time. He would typically work at home for a day or two and then come in to compare notes and exchange a demo or two. When he got his Big Mac, he simply didn't come in or call in for two weeks. Finally I called his home just to check that he was OK, and found him nearly expired. As I recall he said something like, although we had built this incredible system in which you could literally change anything and keep running, until that time processor speed and screen real estate had kept our ambitions in check. Now, at a single stroke, those barriers had simply vanished, and we could just go on doing one thing after another with nothing to hold us back except, ultimately, our ability to stay awake.
- Dan
Prof. Dr. Stéphane DUCASSE (ducasse@iam.unibe.ch) http://www.iam.unibe.ch/~ducasse/ "if you knew today was your last day on earth, what would you do different? ... especially if, by doing something different, today might not be your last day on earth" Calvin&Hobbes
"The best way to predict the future is to invent it..." Alan Kay.
Open Source Smalltalks: www.squeak.org, www.gnu.org/software/smalltalk/smalltalk.html Free books for Universities at http://www.esug.org/sponsoring/promotionProgram.html Free Online Book at www.iam.unibe.ch/~ducasse/FreeBooks.html
On 3/18/03 4:59 AM, "Stephane Ducasse" ducasse@asterix.unibe.ch wrote:
Hi all
Thanks for all these notes. While reading them, I was thinking that they could make a really fascinating book. Dealer of lightning is good but a book really focused on the Smalltalk pionners would be really interesting. May be collecting all the material into a dedicated wiki would be a first step.
Stef
I know that inventing the future is more fun that chatting about the past but still...
I send a private message to Hannes say something similar. To all pioneers and tale / legend inspirers, share with us that days. I sure someone collect all in a book.
Edgar
Does anyone happen to know how many message sends per second one could get on a Dorado, or on any of these early machines? Or the numbers from any of the standard benchmarks that one could run today?
I was just wondering how many (decimal) orders of magnitude we have gained since those days, not so very long ago ...
Andrew
Does anyone happen to know how many message sends per second one could get on a Dorado, or on any of these early machines? Or the numbers from any of the standard benchmarks that one could run today?
I was just wondering how many (decimal) orders of magnitude we have gained since those days, not so very long ago ...
The Dorado was quite a machine for a personal computer -- the size of a small refrigerator and packing ECL circuitry with a micro cycle time of 25 nanoseconds. We predicted that it would get us under a microsecond per bytecode, and it did -- just. I don't remember the send speed, but it should have scaled pretty much as now, relative to the bytecode speed, which would put it around 40K sends per second.
These speeds are roughly 1/50th the speed of my Mac Powerbook. As I recall, the size of main memory was also around 1/50th what I have now, and the disk space around 1/1000th what one expects in a decent laptop today.
... and we were breathless in the face of such performance ;-)
- Dan
At 22:10 -0700 04/21/2003, Dan Ingalls wrote:
...
These speeds are roughly 1/50th the speed of my Mac Powerbook. As I recall, the size of main memory was also around 1/50th what I have now, and the disk space around 1/1000th what one expects in a decent laptop today.
... and we were breathless in the face of such performance ;-)
As well you should have been! A factor of only 50 behind where we are 20 years later says a lot about just what a remarkable machine the Dorado was for its time. As those of us who couldn't get hold of one still remember.
Allen W-B or anyone else: do you happen to have comparable numbers for the Magnolia?
Andrew
Hi Allen --
Your presumption below is correct. We did start with version 1.
Cheers,
Alan
----
At 11:03 AM -0800 3/17/03, Allen Wirfs-Brock wrote:
At 10:30 AM 3/17/2003 -0800, Tim Rowledge wrote:
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
Almost none and very few. The Magnolia was a predecessor of the commercial 4404 which had a massive 1Mb of ram. And a stunningly fast 8MHz 68000 cpu.
Actually, the 4404 (which was announced in July 1984 and shipped in Jan 85) had a 10Mhz 68010 (it had a virtual memory OS which plan 68000s did not support). It was available in 1 and 2 MB configurations. Smalltalk ran ok in 1 MB but for "large" projects 2MB was preferable.
Tektronix Smalltalk was based upon Smalltalk-80 Version 2 and my recollection was that the initial Tektronix release had about 250 classes.
I presume that Squeak was based upon Smalltalk-80 Version 1 as that was what was originally licensed to Apple (along with Tek, Dec, and HP). As far as I could tell, the primary reason for ST-80 V2 was the Xerox realized that the V1 license terms were, from their perspective, way too favorable to the licensees. V2 provided an opportunity for them to revise the license. There were a number of times at Tektronix that we regretted having updated to the V2 image and license.
Allen Wirfs-Brock
--
Because on the pilot, even when there is 2, 4 or even 32 MB of RAM, you can't get at it in continuous blocks. I think only in 32K blocks. (IIRC) You *could* do a full-scale Smalltalk on a Palm, but you'd be doing a lot more than a port of Squeak, you'd have to write a new VM to account for the way PalmOS works.
Regards, Aaron
Aaron Reichow :: UMD ACM Pres :: http://www.d.umn.edu/~reic0024/ "a system based on exchanging products inevitably channels wealth to a few, and no governmental change will ever be able to correct that." :: daniel quinn
On Mon, 17 Mar 2003, Tim Rowledge wrote:
Hannes Hirzel hannes.hirzel.squeaklist@bluewin.ch wrote:
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
Almost none and very few. The Magnolia was a predecessor of the commercial 4404 which had a massive 1Mb of ram. And a stunningly fast 8MHz 68000 cpu.
So, why was it that we couldn't do a fullscale Smalltalk for a palm pilot? Errrrr.....
tim
-- Tim Rowledge, tim@sumeru.stanford.edu, http://sumeru.stanford.edu/tim Strange OpCodes: SDS: Sort of Do Something
On Mon, 17 Mar 2003 13:22:52 -0600 (CST), Aaron J Reichow reic0024@d.umn.edu wrote:
Because on the pilot, even when there is 2, 4 or even 32 MB of RAM, you can't get at it in continuous blocks. I think only in 32K blocks. (IIRC) You *could* do a full-scale Smalltalk on a Palm, but you'd be doing a lot more than a port of Squeak, you'd have to write a new VM to account for the way PalmOS works.
In fact, there is a full-scale Smalltalk on the Palm -- it is called "Pocket Smalltalk", and the website is:
http://www.pocketsmalltalk.com
It is not supported by the original author, nor by the person who took over development and set up the web site, but there is a Yahoo Groups email list that has a few people on it that use it regularly and answer questions.
Later, Jon
-------------------------------------------------------------- Jon Hylands Jon@huv.com http://www.huv.com/jon
Project: Micro Seeker (Micro Autonomous Underwater Vehicle) http://www.huv.com
I don't know if I'd call PocketSmalltalk a "full-scale" Smalltalk- you can't do development in PST *on* the Palm device itself.
Regards, Aaron
Aaron Reichow :: UMD ACM Pres :: http://www.d.umn.edu/~reic0024/ "the end of the human race will be that it will eventually die of civilization. " :: r. w. emerson
On Mon, 17 Mar 2003, Jon Hylands wrote:
On Mon, 17 Mar 2003 13:22:52 -0600 (CST), Aaron J Reichow reic0024@d.umn.edu wrote:
Because on the pilot, even when there is 2, 4 or even 32 MB of RAM, you can't get at it in continuous blocks. I think only in 32K blocks. (IIRC) You *could* do a full-scale Smalltalk on a Palm, but you'd be doing a lot more than a port of Squeak, you'd have to write a new VM to account for the way PalmOS works.
In fact, there is a full-scale Smalltalk on the Palm -- it is called "Pocket Smalltalk", and the website is:
http://www.pocketsmalltalk.com
It is not supported by the original author, nor by the person who took over development and set up the web site, but there is a Yahoo Groups email list that has a few people on it that use it regularly and answer questions.
Later, Jon
Jon Hylands Jon@huv.com http://www.huv.com/jon
Project: Micro Seeker (Micro Autonomous Underwater Vehicle) http://www.huv.com
On Mon, 17 Mar 2003 13:54:50 -0600 (CST), Aaron J Reichow reic0024@d.umn.edu wrote:
I don't know if I'd call PocketSmalltalk a "full-scale" Smalltalk- you can't do development in PST *on* the Palm device itself.
Well, I guess you could look at it that way... I was thinking more along the lines of having a proper VM, with an interpreter and garbage collector, and a reasonably complete class hierarchy.
I think it would be possible to extend Pocket Smalltalk to include a compiler in the system, and the ability to save the image on the device after starting it. Probably not trivial, but it should be doable.
Later, Jon
-------------------------------------------------------------- Jon Hylands Jon@huv.com http://www.huv.com/jon
Project: Micro Seeker (Micro Autonomous Underwater Vehicle) http://www.huv.com
Aaron J Reichow reic0024@d.umn.edu wrote:
Because on the pilot, even when there is 2, 4 or even 32 MB of RAM, you can't get at it in continuous blocks. I think only in 32K blocks. (IIRC) You *could* do a full-scale Smalltalk on a Palm, but you'd be doing a lot more than a port of Squeak, you'd have to write a new VM to account for the way PalmOS works.
Actually, now that I think of it again I recall being told not too long ago that there is in fact a way to get 'below' the memory structure somehow such that it seemed likely that to be possible to do a sensible port. Unfortunately the HandSpring guy that told me about this has now moved on somewhere else and I can't find him.
My intended point was that we seem to have forgotten how to manage with tiny machines. I'm sitting here at a 200MHz high performance descendant of the feeble 8MHz ARM cpu I used for the Active Book and the interactive performance is nowhere near as good. Ok, we're running 16bit colour instead of monochrome and the Squeak vm is not as clever as the Brouhaha one, but it really ought to be at least as good.
tim
Tim Rowledge wrote:
My intended point was that we seem to have forgotten how to manage with tiny machines. I'm sitting here at a 200MHz high performance descendant of the feeble 8MHz ARM cpu I used for the Active Book and the interactive performance is nowhere near as good. Ok, we're running 16bit colour instead of monochrome and the Squeak vm is not as clever as the Brouhaha one, but it really ought to be at least as good.
I've always thought this is a somewhat suprising general trend.
I understand we ask our computers to do more than we used to, but sometimes you'll be doing a simple task, say a search and replace on a large text file using some random utility on some random OS, and a newer machine will take as long or longer than (the memory of) a machine from a decade ago.
I remember being at MacWorld '97 and Apple touting their new "live scrolling" Finder windows, something NeXT had been doing for nearly a decade and countless others before that, and now Apple has come full cycle back to a prettier NeXT OS, and somehow made a 1GHz G4 feel no more responsive than a 25MHz 68040 (unless you're playing video).
Somewhere, collectively, we've forgotten how we accomplished the impossible under limited resources. Even projects like uClinux needs nearly a meg of RAM for the kernel, but what would we have built with the same hardware ( http://www.uclinux.org/ucsimm/ ) two decades ago? (Or 25? Am I really that old?) It's as if we're dazzled by the myriad new possibilities and forget that the dreams of 20 years ago are still a stone's throw better than the status quo today.
The average embedded system has the capability of an enviable workstation from twenty years past. Give me my $30 Lisa Palmtop. I want to see if I can fill up a 16MB Flash.
Or better yet, let's replace all those TI85's in our kids' backpacks with the Pocket Mac Classic. I suspect the graphing calculator will still work, and maybe ClarisWorks will make a revival.
-Jack
On Tue, 18 Mar 2003 09:28:16 -0800, Jack Johnson fragment@nas.com wrote:
I remember being at MacWorld '97 and Apple touting their new "live scrolling" Finder windows, something NeXT had been doing for nearly a decade and countless others before that, and now Apple has come full cycle back to a prettier NeXT OS, and somehow made a 1GHz G4 feel no more responsive than a 25MHz 68040 (unless you're playing video).
I think there are a few things that cause this:
Back in the mid 80's, we were lucky to have a machine that had 8 bit color with a 640 x 480 screen. That's about 300 KB of screen data to move around. If you had a mono machine, like the Macs were, 640 x 480 is only 38 KB of data.
Now, I run a 1600 x 1200 desktop, with 32 bit color. That's a factor of 25x more data to move around. When you compare it to the mono 640 x 480, its a factor of 200x.
Second, I think application programmers are lazier now because they can be. If you look at state of the art games in 1985 and now, the new games are breathtaking in what they do. Game developers have always been pushing the envelope, and back in the 80's and early 90's games were basically written to run on the metal. Now, most games (at least in the Windows world) run as proper Windows applications, using services provided by the OS, and they are definitely orders of magnitude better/faster (from a purely technical perspective) than what we had 15 years ago.
Even game developers are getting "lazy", because most game engines are written in C or C++, and the gameplay itself is often scripted in some higher-level, interpreted language. Nobody writes game engines in assembler anymore.
If application developers spent as much time on algorithm design and tuning as game developers do, our applications would scream. But, the simple reality of the situation is, we don't.
Another factor is we're doing more now than we did 15 years ago. Right now, I think nothing of scanning a 4" x 6" photo on my wife's scanner at 600 dpi, with 48 bit color. That makes a TIFF file that is about 50 MB. I can transfer that file to my laptop through a router over a self-configuring wireless 802.11 card I can plug in and unplug anytime, in less than a minute. I can open that file in a graphics editor, and not even think about whether I have enough memory (half a gig on my laptop).
Another example -- back in the mid-80s, when Pixar did the first version of Luxo Jr, I think they said it took something like 3 hours per frame to render it. With a new consumer level graphics card like a GeForce 4 that I can buy for a few hundred bucks, you can render that movie in real-time, complete with totally dynamic lighting.
Anyways, I'll end my rant here...
Later, Jon
-------------------------------------------------------------- Jon Hylands Jon@huv.com http://www.huv.com/jon
Project: Micro Seeker (Micro Autonomous Underwater Vehicle) http://www.huv.com
I hate to even mention this, but live scrolling windows in Smalltalk were one of the demo items that Dan Ingalls showed Steve Jobs when he came to visit PARC for the first time in '79, and they were as responsive or more than they are now. In fact, one of the fun on the fly changes that Dan made was in response to a question from Steve as to why the scrolling had to be line by line. Dan popped up the browser and made it go scan line by scan line (with very acceptable performance) in just a few seconds of changes. Needless to say, the Apple visitors had never seen any like this.
No one "of a certain age" has forgetten how to do this stuff. It's really a question of what amount of work should need to be put into optimization in an experimental system that is trying to change. The really annoying thing to me is Chuck Thacker's and Butler Lampson's estimation that today's CPUs (with stuff they connect to, etc.) are about 1000 times less efficient than was the Alto (whose 30th birthday is in the first week of April). This is what is killing experimentation. The choices Apple and other companies made as to what kinds of CPU resources they should try to use have crippled the industry (and was one of the main reasons why Smalltalk wasn't a factor in the early 80s when people's opinions and habits about programming were forming.
Cheers,
Alan
-------
At 9:28 AM -0800 3/18/03, Jack Johnson wrote:
Tim Rowledge wrote:
My intended point was that we seem to have forgotten how to manage with tiny machines. I'm sitting here at a 200MHz high performance descendant of the feeble 8MHz ARM cpu I used for the Active Book and the interactive performance is nowhere near as good. Ok, we're running 16bit colour instead of monochrome and the Squeak vm is not as clever as the Brouhaha one, but it really ought to be at least as good.
I've always thought this is a somewhat suprising general trend.
I understand we ask our computers to do more than we used to, but sometimes you'll be doing a simple task, say a search and replace on a large text file using some random utility on some random OS, and a newer machine will take as long or longer than (the memory of) a machine from a decade ago.
I remember being at MacWorld '97 and Apple touting their new "live scrolling" Finder windows, something NeXT had been doing for nearly a decade and countless others before that, and now Apple has come full cycle back to a prettier NeXT OS, and somehow made a 1GHz G4 feel no more responsive than a 25MHz 68040 (unless you're playing video).
Somewhere, collectively, we've forgotten how we accomplished the impossible under limited resources. Even projects like uClinux needs nearly a meg of RAM for the kernel, but what would we have built with the same hardware ( http://www.uclinux.org/ucsimm/ ) two decades ago? (Or 25? Am I really that old?) It's as if we're dazzled by the myriad new possibilities and forget that the dreams of 20 years ago are still a stone's throw better than the status quo today.
The average embedded system has the capability of an enviable workstation from twenty years past. Give me my $30 Lisa Palmtop. I want to see if I can fill up a 16MB Flash.
Or better yet, let's replace all those TI85's in our kids' backpacks with the Pocket Mac Classic. I suspect the graphing calculator will still work, and maybe ClarisWorks will make a revival.
-Jack
--
Alan Kay wrote:
No one "of a certain age" has forgetten how to do this stuff. It's really a question of what amount of work should need to be put into optimization in an experimental system that is trying to change. The really annoying thing to me is Chuck Thacker's and Butler Lampson's estimation that today's CPUs (with stuff they connect to, etc.) are about 1000 times less efficient than was the Alto (whose 30th birthday is in the first week of April). This is what is killing experimentation.
Part of me also wonders what will happen if/when we see multiple terahertz machines and beyond.
I think some of the stagnation we see arises from the idea that the next generation of hardware will solve any performance problems. When we reach a point where the next generation is twice as fast as "nearly instantaneous" and we push the limits of human perception, we may (collectively) start to go back and say, "OK, we can't do it qualitatively faster, how can we do it better?" Hopefully, that might be an impetus to go back and look at all the work that's been done before (and being done now) in a new light.
-Jack
P.S. If so, it's interesting how limited resources and limitless resources could converge to produce similar trends.
On Mon, Mar 17, 2003 at 01:22:52PM -0600, Aaron J Reichow wrote:
Because on the pilot, even when there is 2, 4 or even 32 MB of RAM, you can't get at it in continuous blocks. I think only in 32K blocks. (IIRC) You *could* do a full-scale Smalltalk on a Palm, but you'd be doing a lot more than a port of Squeak, you'd have to write a new VM to account for the way PalmOS works.
Since PalmOS 3.5 you may use FtrPtrNew/FtrPtrFree to allocate blocks over 64K. The only inconvenient is that access is a bit slower and that you have to use DmWrite.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Monday 17 March 2003 12:22 pm, Aaron J Reichow wrote:
Because on the pilot, even when there is 2, 4 or even 32 MB of RAM, you can't get at it in continuous blocks. I think only in 32K blocks. (IIRC) You *could* do a full-scale Smalltalk on a Palm, but you'd be doing a lot more than a port of Squeak, you'd have to write a new VM to account for the way PalmOS works.
Regards, Aaron
There already is a SmallTalk for Palm, called PocketSmalltalk... http://www.pocketsmalltalk.com; It's free as well and they are working on the port for WinCE as well.
Brian
Also, check out -- I think it is described in the Green Book edited by Glen Krasner -- the Smalltalk-78 system done by Dan Ingalls for the 86 processors we had in the Notetaker at PARC. IIRC, this only had about 6-8K of actual machine code for the lowest kernal of the VM and ran the whole system in a fairly small amount of RAM (no hard disk on this machine).
Cheers,
Alan
At 7:02 PM +0100 3/17/03, Hannes Hirzel wrote:
merlyn@stonehenge.com (Randal L. Schwartz) wrote:
> "Joshua" == Joshua 'Schwa' Gargus schwa@cc.gatech.edu writes:
Joshua> Apple was one of the original licensees of Smalltalk-80 when Joshua> it was first release outside of Xerox PARC.
As was Tektronix. I first used Smalltalk on a Magnolia (68K) at Tek in 1982.
Just curious: How much RAM did that system have and roughly how many classes did the image have you were working with?
-- Hannes
--
squeak-dev@lists.squeakfoundation.org