The Epson HX-20 – A Contrarian's View (2021) | Hacker News Reader

The Epson HX-20 – A Contrarian's View (2021)

feertech.com

65 pointsrbanffy52 comments

My first sizable programming job in high school was with one of these, I think. A special education teacher had written an app to create a Blissymbolics <https://www.blissymbolics.org> communicator and hired me to tune it up, because some of his more prolific students had mastered more symbols than the app could readily handle.

I spent quite some time with the machine, and felt it was fairly capable for the time (my reference being CBM a.k.a. PET desktops).

"... the charging circuit is, to put it mildly, primitive. There is no charging indicator. Nor is there a sign that the batteries are fully charged. The HX-20 manual states that empty batteries should be put on charge for 8 hours. Leaving them longer risks damaging them, but without any way to tell that they are full there is no way to be sure you’ve timed it right. "

This is normal for NiCd and later NiMH batteries. They're recharged applying regulated (ideally constant) current for a given time, and it's impossible to get a figure of their charge status just by sampling their voltage like with Lithium cells, which are dangerous indeed if abused, still much much easier to recharge. The only known ways to check those cells status are to either continuously sample the energy put in, or to intercept the so called delta V and delta T, a very slight (millivolts) decrease in cell voltage near the end of charge, paired with an increase of temperature which indicates the cell can't store more energy. This becomes even more difficult with battery packs, and makes measuring their status very hard when charging, and completely impossible if kept in use at the same time. Many portable safety anti-blackout lamps from the 70s to the 90s used NiCd and NiMH batteries, and many of them broke after some time because they kept their batteries under constant charge.

The HX-20 lost while the Tandy Model 100 won, because the latter found a real market: Journalists and others who wanted to write on the road. The former's screen isn't large enough for writing despite the good keyboard, and vertical markets where the screen size didn't matter as much weren't large enough to drive HX-20 sales. BYTE's review explains this. <https://archive.org/stream/byte-magazine-1983-09/1983_09_BYT...>

This is a weird take - dunking on a product from more than forty years ago because it doesn't have modern ease of use.

It's like complaining a 1990s dumb phone wasn't an iPhone, or that a 1920s string and canvas biplane wasn't a 1960s SR-71.

Yes, obviously, but why would anyone expect that?

Indeed, another instance of presentism...

>This is a weird take - dunking on a product from more than forty years ago because it doesn't have modern ease of use.

Oh, good grief. The author repeatedly discusses the limitations of the product as they would have been seen by someone today and 40 years ago.

A question I have about this era of of computing. Why did DOS win?

BASIC, for all it's faults, would have been a much nicer shell environment than the extremely limited CPM/DOS syntax. From this modern perspective I would have guessed at the time that BASIC would have grown better file management operations and been the home computers equivalent to the unix shell. that is, not a great programing language but let you run commands and easily, using the same language, program them.

Instead the primitive(in comparison to BASIC syntax) CPM command interrupter and moreso Microsofts "we have CPM at home" DOS syntax won the home computer market. why? every computer in that critical 1981 lineup, including the IBM PC, had a BASIC rom.

This may come as a shock to someone reading HN, but approximately 0.0% (rounded) of all people care about shell programming.

Whatever gets programs to run the easiest wins. DOS was easier than BASIC.

Hell, I had been using computers for well over a decade to accomplish tasks when they plopped me down in front of a Sun SPARCStation 5 and someone mentioned that the Bourne Shell could automate a bunch of the stuff I was doing.

Because BASIC is a programming language (and was included with DOS, cf MS BASIC)

DOS won because it offered a consistent way to manage your files and applications, it was "better" than CPM, and it was available on more machines than anything else. (DOS didn't win. The PC won. DOS tagged along)

It's worth noting that it didn't, for a long time, "win" the home computer market. That was wildly diverse, until PCs were good & cheap enough to replace home computers. And at that point, sheer prevalence started squeezing out home machines. That world ended, more or less, with the Atari ST and the Amiga 500 - because both of those were great, and both were niche tools.

PCs were the thing that you could use for games and work stuff. It turns out "getting stuff done" matters much more to people than engaging in purity debates.

(Also relevant in this context: https://www.dreamsongs.com/RiseOfWorseIsBetter.html)

I am deeply steeped in the history of computers and the biggest three things I can point to as the reason (MS-)DOS won are:

- Licensing: Most computers either had custom operating systems that were not shared with other hardware vendors, or in the case of BASIC frequently, were licensed themselves.

- IBM letting the genie out: The BIOS on the IBM PC 5150 was cloned, quickly and legally, and other companies started making compatibles. This caused an explosion of computer variety in a few short years for a single platform.

- Microsoft: DOS usually means "Microsoft DOS", Microsoft also was responsible for many of the BASIC environments of early systems as well. The ability to buy your OS from someone else lowered the pressure on hardware makers. IBM also favoured Micorsoft's DOS over CP/M-86 and stopped supporting it quickly.

All this meant the PC compatible ecosystem with Microsoft DOS became easy to make from a hardware side, and lacked a single point of failure like Apple, Radio Shack, Commodore. Atari, etc. There were other MS-DOS compatible DOS's out there, but MS-DOS was usually the one shipped with computers to be as "IBM compatible" as they possibly could and gained dominance through that.

EDIT: To those who may not be aware, BASIC did become more OS like before going away. HP BASIC was extremely feature packed before HP-UX replaced it and was more capable than MS-DOS in many ways. It evolved far beyond just a programming language.

> This caused an explosion of computer variety in a few short years for a single platform.

The impact of this point can not be overstated. 99% of businesses make a much larger investment in software (and people!) than hardware. The idea that compatible hardware systems existed was a great hedge on their investment in software and training. For most businesses, this would be a no-brainer!

Over a short time, other propietary/non-compatible systems were relegated to home use, education, and gaming.

First, IBM created an open standard which stuck due to their size.

Then the killer app Lotus 123 came out and it didn't use Basic and nothing you wrote in Basic could compete with 123 either.

My older brother bought a $4000 TI Professional computer, added 123 and made enough to pay for it over 1 weekend (time sensitive tax related documents for a real estate transaction). He couldn't have done it with Basic.

This and piracy was the kicker.

And the business machines companies paid $$$ for to get access to 123 ended up at home a few years later as newer machines entered the workforce.

DOS won because IBM PCs and PC compatibles won. Most of the popular earlier 8 bitters had a merged REPL that did both BASIC and 'operating system' interaction. So, for instance, on an Apple ][, the default prompt put you in the BASIC REPL from which you could also run the Apple DOS (later ProDOS, neither related to MS DOS) commands.

True. DOS wasn't better, it's just the first OS I've used, but CP/M works just fine and is easy to learn. Not that that's relevant to this machine, it has no real OS, just a basic system for managing the hardware.

Because DOS is a disk OS and BASIC is an interpreted programming language.

The point of DOS isn't just the command line. It provides a set of standard abstracted system calls that can be used by any developer working in any language with a compatible compiler, including assembler, to create any application.

A command line interpreter sits on top of the library for immediate use, but that's just the part the user sees. The meat of DOS is the function library.

BASIC doesn't work like that. It is an application, and other software can't access its commands from outside. It was never designed to allow that, and extending to it to make it possible would have created a confusing, slow, inefficient, and unreliable mess.

Still, many 8-bit micros extended BASIC to provide a "shell" for their DOS equivalent. For example, the Apple II with both ProDOS and DOS 3.3. There was still an underlying DOS library. It would've made sense for the PC to do the same, given all the contemporary 8-bit systems.

Why DOS? 1) Gary Killdall thought too high and mighty of CP/M (maybe rightly) while 2) Bill Gates had nothing to lose from selling IBM what they needed to the new PC. Gates told them what they wanted to hear, and he won. That dweeb was smart and his dad being a lawyer taught him about contracts. He then sold DOS to all the clones with a forced per CPU license, even if MSDos wasn't sold with the Clone. Gates made bank early.

Interesting side note. It took years for PC users to gain multi-user capabilities, mostly when with Novell connected machines and brought file/record locking etc. And before the PC multiuser was already in MP/M and Unix on machines like Altos and Molecular Computer. Killdall would have probably got eht market back if he hadn't died.

And fast forward today, deep down Windows still runs much DOS code. Like Veeger it's morphed into Windows whatever.

That was strongly true for a long time through Windows 9*, but with Windows NT/2000 much less so as the core was primarily influenced by David Cutler's experiences with VMS.

There’s no DOS code running deep down in Windows. What do you have to gain by saying these things?

> There’s no DOS code running deep down in Windows.

Correct.

> What do you have to gain by saying these things?

Stop right there. From the site guidelines: Assume good faith. You've been here long enough that you should know that.

When someone says something wrong, don't immediately assume that they say wrong things because they have something to gain. Don't accuse them of having something to gain. People can just be mistaken.

So correct, but don't accuse.

The question was not dos vs cpm(both of which are superficially the same thing), but dos vs basic, both provided by Microsoft. People get tripped up with basic being a programing language, but imagine dos only instead of the braindead dos command syntax you had a full basic repl as a command line. Note that good dos(with directories, environment variables, and redirection came much later)

Did microsoft not provide disk and program launching functions in basic? Perhaps dos used less memory, memory was a very very tight commodity in 1981.

To be honest, I don't think BASIC makes a good command line. It's been done but it's not really that great.

As someone who computed in the DOS era, programmed in BASIC, and even installed different command interpreters with more features -- I never once thought that BASIC would have made this better.

The other factor was actually disk and memory size -- you want to keep things as minimal as possible to keep things running fast. When your command line interpreter is reloaded from floppy disk every time you return from a program, you want that to be fast.

One reason is limited memory and address space- the in-memory size of CP/M was something like 9K (including BIOS, BDOS and CCP). BASIC is at least another 8K. If you are using CP/M as a program loader, you don't want to waste that 8K.

Another is that BASIC was not extensible, at least not in any standard way. But CP/M was: to add a command to CP/M, you just add a .COM file of that name to the disk.

IMO it's that final point that makes the difference - the fact it was a ROM. If your system has BASIC built-in, you have the version of BASIC that was built into it. If your system has DOS built-in, you can run whatever version of BASIC suits you best, and as time goes by you can even upgrade to newer versions; maybe you can't upgrade DOS itself, but it's so basic that there's not really anything to upgrade (the only parts you might want to are things like drivers, and those are tied to your hardware anyway). Time and again systems that move more into "userspace" have won out over systems that implement complex functionality in "system space".

Epson is one of those companies that creates a solid product and just lets it ride. Like their V600 scanner has been $250-350 for its entire lifetime and is still the best on the market; it came out in 2009.

I remember my family having bought a PC and an Epson printer sometime in the 1990s. Both were roughly in the ~$1000 range. Almost a decade later we considered selling the computer for ~$50, and gave it away instead. We sold the printer for about the same price we bought it. (It was a type of printer that was in demand as small companies could print invoices and receipts with it).

EDIT: I checked the model and incredibly it is still sold, between $300-$1000 on ebay, etc.

Scanners are also a tech that has pretty much plateaued - correct me if I am wrong, but the linear arrays are fairly cheap and the resolution is a product of the linear array and the minimum step size, and steppers have not gotten much better. There are 2x+ more dense linear arrays than the v600 has that exist, but that only would improve one dimension of the resolution.

Also, the v600 is a great consumer scanner, but OH BOY can you get more fancy, for film you have the hasselblad/imacon line that are still a few thousand used, then you have drum scanners for the highest possible film scanning quality. For a more direct v600 comparison, there is this really, really old scanner that outperforms it: https://www.scansolutionsonline.com/media/1168/cezanne-elite...

A more modern flatbed scanner: https://epson.com/For-Work/Scanners/Photo-and-Graphics/Epson... Sadly lower res than what used to be out there.

The Imacon Flextight has been discontinued for a while now. Flatbed scannners are stagnant because there isn't enough demand to drive investment. Sheet-fed document scanners are still lively, but quality has gone down, my 2010 vintage ScanSnap S1500M has a high-quality CCD sensor, but newer models are almost invariably CIS with much lower color fidelity for scanning artwork.

High-quality XY prepress scanners like the Cézanne you mention, or prepress leaders like Fuji Lanovia, Creo-Scitex are also discontinued since the industry's gone fully digital and doesn't expect to scan medium-format slides any more.

No wonder Gen-Z finds scanners exotic and terrifying:

https://www.theguardian.com/technology/2023/feb/27/gen-z-tec...

The only actively produced scanner that can outperform the V600 is Epson's V850.

Fair enough - personally, I'd use a digital camera to scan large things anyway, possibly using an XY gantry if needed. Use a monochrome sensor + color filter wheel for extra points.

Scanners terrify generations equally in my opinion ;)

I've owned flatbeds since the original SCSI CanoScan 300 in 1994 (couldn't afford the 600dpi CanoScan 600 then), so I'm jaded. The digital camera option works, but only if you have a high-CRI light source, a proper macro lens and a good copy stand. The Fujitsu SV600 is an excellent option as well.

Museums actually used scanning large-format backs like the Better Light Dicomed, a flatbed scanner-like back that slots into a 4x5" large-format camera, so the camera is also a scanner...

Still? I would have thought by now they would have switched to something like a Phase One IXH - https://heritage-digitaltransitions.com/dt-ixh-150mp-cultura...

I have a broken betterlight I got just to put on my shelf haha. Scan backs are neat!

What a coincidence, I pulled out my S1500M (vintage 2010-10) to do some scanning a few weeks ago (after several years) and found the rubber rollers had melted and become like superglue. Currently reading about how I might try to do an ad-hoc fix, because forums say disassembly and reassembly is very finicky indeed. Now that I learn it's CCD I am more motivated to repair.

The roller kit is a field-replaceable unit and Fujitsu supplies them in a boxed kit, so it can't be that hard (I'm still on my original roller, even though I bout a roller replacemnt kit, so I have no idea how hard it is).

It's a great unit, but official software support lapsed a few years ago so I am now using it with third-party software (ExactScan Pro).

Here are a few scanners I ran through a comparative color fidelity test a while ago, I was surprised how good it is for scanning artwork when it was designed for documents:

https://blog.majid.info/scanner-group-test/

This thing was pretty solid. The circuitboard is silk screened on both sides with labels for everything, test points, and a trademark Epson thick technical manual. Repair is fairly easy except for the self destructing batteries that take out the FFCs inside. Still, if you can get your hands on one, you can probably repair it. The technical manual even goes into detail on how to install the capacitors, recommending about 1mm below the capacitor for the legs to come up and connect, so in the future if you have to replace them, you just cut the legs off and remove the remnants.

Epson has by far the most complete online collection of documentation I've ever seen. It even has manuals <https://files.support.epson.com/pdf/qx10__/qx10__u1.pdf> for the QX-10 <https://en.wikipedia.org/wiki/Epson_QX-10> (1983), a CP/M machine!

"For a start, Epson insists that it must only be run from batteries"

I don't think that's correct. I had one handed down to me in the early nineties. Afair (well, it's a long time), I operated it w/o batteries at all.

It seems like it would have been easy to use as a battery-powered POS terminal, with just some simple BASIC software. That might not have been sufficiently better than a mechanical cash register or a carbonless-copy receipt pad though to justify it.

You'd think so, but I can't find anyone had actually been using it as such. Since it can read and write tapes, it seems it would be perfectly suited to running a retail store's day to day business. The barcode scanner and some basic database software would have probably sold like hotcakes as a portable POS system.

Were there barcode scanners that were efficient enough to be battery-operated in 1981?

Looks like there was one you had to swipe over the barcode, likely just a single CCD type device https://computarium.lcd.lu/photos/albums/EPSON_HX20/album/sl...

The printer is very slow, one line at a time. It would not be great as a POS terminal, but it could connect to a serial printer over RS-232 and external disks and a screen via the other TTL-level 36400 bps port.

I have one, funnily it was from the German military that used it for artillery calcularions.

Who said there's no digitalization in Germany?

As a forward observer once radioed to artillery during WWII, "Stop computing and start shooting!"

I used one long ago: I remember that, as with the TRS-80 model 100, the keyboard was outstanding.

Yep cool, but the tiny screen was ridiculous and hard see, let alone sell.

Its really nice that the language of that era still informs and inspires designers today.

I have a small collection of ClockworkPi toys, most recently a couple of uConsole's, and they are delightful reminders of an aesthetic that still lingers on, in the minds of those raised on computers-as-life.

The cyberdeck guys are going to continue this, I just know it. I hope we see an Epson HX-20-esque re-imagining of the ClockworkPi somewhere along the line ..

The later Epson PX-8 has a much more usable screen in a similar form factor.

The DevTerm is their closest form factor imo https://www.clockworkpi.com/home-devterm. I bought one for kicks, the keyboard kinda sucks but the built in receipt style printer is badass.

The uClonsole also looks pretty sweet.

If there was a devterm with a 80% keyboard I'd gladly buy it, but that tiny keyboard on it makes it a no from me.

I’d LOVE a Devterm the same size as the TRS-80 model 100 or the HX-20. Would get it immediately.

Reply on news.ycombinator.com