4 ms·

Even though I only scanned the article, many of those computers appear in Turing's cathedral G.Dyson[1], book which I enjoyed, given that there is not a lot of

Even though I only scanned the article, many of those computers appear in Turing's cathedral G.Dyson[1], book which I enjoyed, given that there is not a lot of history lessons at the universities (at least the one I went to in Buenos Aires... UBA). [1] https://openlibrary.org/search?isbn=0375422773 https://openlibrary.org/search?isbn=0375422773

alexanderdmitri 9y ago
I always thought it would be neat to have a CS History lab where every couple of weeks the students would program with a particular generation of computer, moving up through time and getting a true sense of the ingenuity at each step. Might be expensive to house and maintain something like the SSEM or the CSIRAC, but universities have spent money on sillier things.
- drittich 9y ago
  I'd be all over that!
- jerf 9y ago
  Emulators are good enough for most students. I did "circuit design" in an entirely emulated environment and got what I needed out of it as a computer programmer. It would not have been adequate for a real EE, because for instance it offered unlimited fan-out and fan-in, but from my perspective I wasn't losing much to abstract that away.
  - alexanderdmitri 9y ago
    True. Mixing emulation in would probably save a lot of work and resources. I feel like having to do things like manually punch up your programs on cards and manage with tiny amounts of storage would be a great experience and really force students to work towards elegant solutions. I love getting my Dad talking about some of the labs he's worked in over his career and kind of wish I could get a sense of how much has changed over such a short time.
    - jerf 9y ago
      The other nice thing about emulation is that you get to use modern IO hardware, particularly keyboards. I can imagine having fun programming on a Commodore 64; I have a much harder time imagining having fun doing it with an actual Commodore 64. The keyboards were not all that great by modern standards when they were brand new, don't get better with age, and certainly don't get better in a student lab environment :)
    - flyinghamster 9y ago
      At my high school circa 1980, the introductory computer course (on a PDP-11/34 running RSTS/E) had us write our first BASIC programs onto mark-sense cards (the same size and shape as punch cards, but you marked them with a pencil). I recall a field for the line number, another for a keyword (pick from a list), and then the remaining part of the card was for variables and operators. After a couple go-rounds with that, we were then allowed to use the terminals, mostly Visual 200, with a few VT100s, an LA36, and an LA120. The Visual 200 wasn't a particularly good terminal; it seems there was always at least one out of order at any given time. It's kind of scary that a $5 Raspberry Pi Zero is much faster and has vastly more RAM than the 11/34, which cost well north of $100K back then. Not to mention that there was a grand total of 28 MB of storage on that 11/34 in the form of two RK07 disk packs - you could fit thousands of RK07 images on a single 64 GB MicroSD card.
- VLM 9y ago
  At least from an EE perspective its interesting to contemplate "here's the highest performance discrete RTL technology that could be designed in 1960 using 1960 transistors... You have free reign to design a RTL NAND gate using any 2017 production transistor you want. In the spirit of IBM mainframes, no power limit. Can your NAND gate break 150 MHz? Can your NAND gate break 15 GHz?" "You can buy FETs that have 1970s performance specs in 2017, very cheaply in fact, but in 2017 you can also buy FETs that output many watts of power at 10 GHz x-band microwave freqs. Here's a DEC flip-chip module containing two flipflops that topped out around 30 or so MHz in 1965, your assignment is to use modern transistors in the lab to make a modern work-alike operate over 10 GHz"
  - pjc50 9y ago
    At that speed the simple fact of using discrete components kills it, surely? Multiple picofarads of parasitic capacitance on the leads, plus gate capacitance itself? > buy FETs that output many watts of power at 10 GHz x-band microwave freqs. Part number? Gate drive for that must be 'interesting'.
- mschaef 9y ago
  Agree this this is a good idea, although not necessarily on real hardware. There's enough both CapEx and OpEx that would be necessary for the hardware that it would mainly tend to restrict access to the course. That said, on emulation, it's a great idea. It's easy to forget how many of the assumptions of C-like languages and Unix-like operating systems are built into the hardware, particularly when there are other ways to go about the task. There are reasons the industry has gone the way it has, but it's worth keeping those ideas around for posterity.

alexanderdmitri 9y ago

I always thought it would be neat to have a CS History lab where every couple of weeks the students would program with a particular generation of computer, moving up through time and getting a true sense of the ingenuity at each step. Might be expensive to house and maintain something like the SSEM or the CSIRAC, but universities have spent money on sillier things.

drittich 9y ago
I'd be all over that!
jerf 9y ago
Emulators are good enough for most students. I did "circuit design" in an entirely emulated environment and got what I needed out of it as a computer programmer. It would not have been adequate for a real EE, because for instance it offered unlimited fan-out and fan-in, but from my perspective I wasn't losing much to abstract that away.
- alexanderdmitri 9y ago
  True. Mixing emulation in would probably save a lot of work and resources. I feel like having to do things like manually punch up your programs on cards and manage with tiny amounts of storage would be a great experience and really force students to work towards elegant solutions. I love getting my Dad talking about some of the labs he's worked in over his career and kind of wish I could get a sense of how much has changed over such a short time.
  - jerf 9y ago
    The other nice thing about emulation is that you get to use modern IO hardware, particularly keyboards. I can imagine having fun programming on a Commodore 64; I have a much harder time imagining having fun doing it with an actual Commodore 64. The keyboards were not all that great by modern standards when they were brand new, don't get better with age, and certainly don't get better in a student lab environment :)
  - flyinghamster 9y ago
    At my high school circa 1980, the introductory computer course (on a PDP-11/34 running RSTS/E) had us write our first BASIC programs onto mark-sense cards (the same size and shape as punch cards, but you marked them with a pencil). I recall a field for the line number, another for a keyword (pick from a list), and then the remaining part of the card was for variables and operators. After a couple go-rounds with that, we were then allowed to use the terminals, mostly Visual 200, with a few VT100s, an LA36, and an LA120. The Visual 200 wasn't a particularly good terminal; it seems there was always at least one out of order at any given time. It's kind of scary that a $5 Raspberry Pi Zero is much faster and has vastly more RAM than the 11/34, which cost well north of $100K back then. Not to mention that there was a grand total of 28 MB of storage on that 11/34 in the form of two RK07 disk packs - you could fit thousands of RK07 images on a single 64 GB MicroSD card.
VLM 9y ago
At least from an EE perspective its interesting to contemplate "here's the highest performance discrete RTL technology that could be designed in 1960 using 1960 transistors... You have free reign to design a RTL NAND gate using any 2017 production transistor you want. In the spirit of IBM mainframes, no power limit. Can your NAND gate break 150 MHz? Can your NAND gate break 15 GHz?" "You can buy FETs that have 1970s performance specs in 2017, very cheaply in fact, but in 2017 you can also buy FETs that output many watts of power at 10 GHz x-band microwave freqs. Here's a DEC flip-chip module containing two flipflops that topped out around 30 or so MHz in 1965, your assignment is to use modern transistors in the lab to make a modern work-alike operate over 10 GHz"
- pjc50 9y ago
  At that speed the simple fact of using discrete components kills it, surely? Multiple picofarads of parasitic capacitance on the leads, plus gate capacitance itself? > buy FETs that output many watts of power at 10 GHz x-band microwave freqs. Part number? Gate drive for that must be 'interesting'.
mschaef 9y ago
Agree this this is a good idea, although not necessarily on real hardware. There's enough both CapEx and OpEx that would be necessary for the hardware that it would mainly tend to restrict access to the course. That said, on emulation, it's a great idea. It's easy to forget how many of the assumptions of C-like languages and Unix-like operating systems are built into the hardware, particularly when there are other ways to go about the task. There are reasons the industry has gone the way it has, but it's worth keeping those ideas around for posterity.