Hacker News Reader

Comment by creshal | Hacker News Reader

Even with an Apple ][ I doubt they could become true heirlooms. Most of them already broke down, and there's no way the survivors will survive 50 years or longer without having all internals refurbished.

Although that's not nearly as bad as this one. SSD, fan, LCD, RAM, battery are all wear parts that won't survive a decade on average, and the rest is only marginally better.

View on HN

Tin whiskers is another problem https://en.m.wikipedia.org/wiki/Whisker_(metallurgy)

Luckily Apple ][ innards are all commodity. Caps, resistors, 74xx parts, some RAMs, some PROMs and a 6502. Everything is replaceable, and there's nothing too crazy.

I was actually all set to say that 6502s were now unobtainable, but it turns out that they're not --- in 40-pin DIP, no less. Assuming they haven't changed the pinout it should just drop right in.

http://www.mouser.co.uk/ProductDetail/Western-Design-Center-...

Mmm, 600nm process!

The 6502 is relatively popular in some embedddd devices. The Tamagotchi used it: http://hackaday.com/2013/05/24/tamagotchi-rom-dump-and-rever...

The trouble with the old architectures like the 6502 and the Z80 is that they were designed before stack-frame languages like C became popular --- which means they don't have any stack-relative addressing modes, which means that accessing stack frames is horribly expensive.

e.g. on the Z80 most compilers have to keep a copy of the stack pointer in an index register, so that they can address the stack; and even then the absence of 16-bit accesses means that you need eight bytes of code and like a billion cycles to read or write a single 16-bit value. (I do know that the Z80 variant in the Game Boy added a ld hl, (sp+offset) instruction, which helps; shame they didn't add a ld (sp+offset), hl instruction...)

The 6502 is even worse; the stack's limited to 256 bytes and isn't relocatable, let alone addressable. The C compilers I've seen maintain their own stack pointer in zero page. They also generate code which at best is terrible and at worst makes you want to claw your own eyes out to make the pain stop.

So:

(a) I'm curious to know what they wrote the Tamagotchi firmware in --- raw machine code? Forth? C? Fortran? (Fortran's actually a good match for these architectures because all variables, including function parameters, are global --- no stack frames required.)

(b) why was it cheaper to use a 6502 core rather than a more modern architecture --- PIC, AVR, MSP430? Licensing? An in-house team of 6502 experts?

That product was made in 1996. Up to then, quite a few gaming devices already used them. Commodore was going bankrupt but company taking over still sold tons of them. So, it was a decent chip, probably cheaper than usual, and might have hired game developers with prior experience in it.

All speculation based on business details at the time. I'd also like to know if it was cheaper or there was a talent advantage in mid-90's.

There's actually a niche market for making pin-compatible replacements for obsolete parts. Mainly for embedded and military industries. For instance, my private research looking for ultra-reliable hardware found RCA 1802 to be a great option:

https://en.wikipedia.org/wiki/RCA_1802

Turns out, it's still around via Intersil with a very thorough datasheet:

http://www.intersil.com/en/products/space-and-harsh-environm...

Another trend is emulating them cycle for cycle and bug for bug with virtualization on modern hardware. Great example is replacing legacy VAXen with NuVAX:

http://logical-co.com/nuvax/

So, there's possibilities for obsolete stuff to keep going. Just can't depend on that happening. Especially if it's a complex and modern SOC rather than simpler chips of the past.

I thought the heirloom part is that a lot of that stuff is user-replaceable on this design, unlike most (all?) other laptops. I assume this is a standard Mini-ITX motherboard?

So you could keep it working with some maintenance over the years.

Its actually an open source hardware design built around a Freescale quad core arm CPU. It is definitely designed to be user replaceable; you can print your own boards if you want.

Info here: https://www.crowdsupply.com/sutajio-kosagi/novena

Designing and making boards for this kind of system is very expensive, especially at small scales. The Novena as a whole sold far too little to really spread out the one-time engineering costs and was very expensive as a result. Only 12 of the Heirloom laptops exist and it's unlikely all 12 owners would buy a replacement board; it's just not going to get made unless one of the owners has the skills to do it themselves and is willing to spend far more than the cost of the laptop to do it.

I'm surprised it turns out to have an ARM processor - I wouldn't have guessed that it would need active cooling. When you consider that many phones would use this same CPU, in a smaller package with less airflow and smaller heatsinks, it seems even more surprising that a fan is required here.

The reason phones can do that is because they throttle heavily after a couple minutes of heavy CPU usage. That wouldn't be very good for a laptop.

The success of the Macbook seems to be arguing the opposite. Most tasks we perform don't require sustained CPU usage.

macbooks are giant passive heat sinks though; a wooden laptop doesn't conduct all that well

But that Freescale CPU isn't open and cost big $$$ to develop. That's the problem: can't use something like that if heirlooms. Only exception maybe is if whole platform is tied to a virtual ISA. That way, hardware might change but SW & interface is the heirloom. System/38 did this to be ported to successively new hardware and interfaces as AS/400, then iSeries, then IBM i.

The ports, though? That looks like USB Type A? I think that'll be replaced by USB Type C sooner or later.

If you think about old computers with their serial ports and floppy drives, it's hard to think these ports will still be in use decades from now.

That's a fair counter but one must remember the cost. The PCB board is usually designed for certain parts. Those parts availability often changes. Replacing them requires redeveloping or implementing them at hundreds of thousands to tens of millions of dollars. It can also be hard to replace a CPU depending on how it's packaged into the board.

So, it would have to be pluggable, open, inexpensive to prototype, and enough volume to spread prototype over users in maintenance.

Reply on news.ycombinator.com