264 comments

[ 3.6 ms ] story [ 327 ms ] thread
Better response times than other websites I browse… maybe they should upgrade to Casios.
I wonder whether it would be fast if it ran some ancient web technology, such as CGI, .Net frameworks for the web or Filemaker web.
I strongly suspect the webpage is entirely static and trying to run any sort of modern serverside framework would make it wail in protest - some simple dynamic elements using the native script might be doable though. It'd be neat to see if they could get a page counter up - just a dumb one that blindly increments the counter whenever any visit occurs.
Maybe we should all skip cloud platfoms and CDNs and buy a bunch of Casios instead.
How many Casios can you fit in a U1 rack?
The calculator's dimensions are: 21.3 x 87.5 x 180.5mm

A 1U rack comes in various depths, let's assume 1m (1000mm), and the width between rails is about 450mm. 1U is around 44mm in height.

So in 1U, you could stack them 2 high, 5 wide, and 5 deep (this leaves a little wiggle room for cables and such), or 50 per 1U, or 2100 in a 42U full rack.

(comment deleted)
we should all upgrade to gemini on casios, not sure if the processor is powerful enough to run the opentls encryption. perhaps a beowulf cluster can be made
Hm, maybe posting this to HN wasn't the best idea...?
Maybe we'll get a follow-up on hosting a distributed cluster of casios to handle global load balancing
c4s?
Nawww, You'd want to call it Casiopeia (sic).
Perhaps vertically scaling this on a Casio FX-CG50PRIZM is an option?
op just need to add some cache servers in front of it and it'll be fine, impressive
Back in my day it was a Beowulf cluster running on hot grits serving Natalie Portman pictures
Oh man, Beowulf clusters, I forgot all about those. How come we never hear about them anymore?
I'm guessing because recurring one-liners don't do as well here as they did on SlashDot.
"Running MPI programs on a cluster of commodity hardware" was perhaps a sufficiently novel idea to warrant a unique name back in the 90s, but these days it's more or less the standard for all but the most exotic and expensive of scientific computing... and even those expensive custom supercomputers tend to still be x86 processors running MPI code.

Basically you don't hear about them any more because it's all Beowulf clusters now.

I think the "Beowulf" name (the name of the first well known computer cluster) was mostly a Slashdot meme. As Slashdot faded, so did the name "Beowulf".
Because now they're just clusters. Lots of people make hobby ones out of rpi's to tinker with
Inferno is available for Nintendo DS, more info at https://gamebrew.org/wiki/Inferno_DS I can verify that it runs, at least, on a Nintendo DSi, though I never actually did anything with it and didn't have another DSi (or enough knowledge) to find out if it could actually network.
I am intrigued by your ideas and would like to subscribe to your newsletter.
In case it completely dies there are archives now available:

http://web.archive.org/web/20210712193555/http://fxip.as2034...

https://archive.is/5a44i

Or the whole thing is small enough to fit in a comment:

  > GET / HTTP/1.1
  > Host: fxip.as203478.net
  > User-Agent: curl/7.76.1
  > Accept: */*
  > 
  * Mark bundle as not supporting multiuse
  * HTTP 1.0, assume close after body
  < HTTP/1.0 200 OK
  < Server: uIP/1.0 http://www.sics.se/~adam/uip/
  < Connection: close
  < Content-type: text/html
  < 
  <!DOCTYPE html>
  <html>
  <head>
   <title>Casio fx-9750GII Webserver</title>
   <link href="https://fonts.googleapis.com/css2?family=Lato&display=swap" rel="stylesheet">
   <style type="text/css">
    * {
     margin: 0;
     padding: 0; 
    }
  
    *:focus 
    {
     outline: none; 
    }
  
    *::-moz-focus-inner 
    {
     border: 0; 
    }
  
    body 
    {
     background: #fafafa;
     font-family: "Lato", sans-serif;
     font-size: 100%; 
     text-align: center;
    }
  
    a 
    {
     color: #0075BF; }
     a:not(:hover):not(:focus):not(:active) {
     text-decoration: none;
    }
  
  </style>
  </head>
  <body>
   <br>
   <h1>Casio fx-9750GII Webserver</h1><br>
   Welcome! 
   This page is hosted on an Casio fx-9750GII graphical calculator, running a SuperH SH4 processor.<br>
   It's running a port of the uIP TCP/IP stack, using SLIP over the 3pin 2.5mm serial port.<br>
   <br>
   <a href="https://github.com/Manawyrm/fxIP" target="_blank">Firmware (uIP port, webserver implementation)</a>
   <br><br>
   <img src="https://screenshot.tbspace.de/vfscewyjzir.jpg" style="width: 700px;"><br>
   <img src="https://screenshot.tbspace.de/ijlnydgsvkw.jpg" style="width: 700px;"><br>
   <i>(these pictures are not hosted on the calculator.)</i>
   <br><br>
  </body>
  </html
So most of the content (images, font) is hosted elsewhere ...
of course just like with most places using content providers :) . It would take forever (relatively)to pull those off the casio vs the html text for the page.
(comment deleted)

    *:focus 
    {
     outline: none; 
    }
  
    *::-moz-focus-inner 
    {
     border: 0; 
    }

Wow, the equivalent of having a building with existing wheelchair ramps and ripping them out because they're "unsightly".

I guess it's a valid decision decision, but pretty crappy for anyone who needs them.

is that to imply that focus outline is meant for accessibility purposes?
Yes, focus rectangles are an essential component of navigating the interface with the keyboard.
Fair. But if a user has accessibility enabled, that should override the css itself no?

Also even with the focus outline disabled on other sites, I am able to navigate with the keyboard to text input fields. OP's site doesn't have any input fields so not sure why they have that CSS in place as I don't think it's doing anything.

You are making the assumption that the user has some modicum of control over the browser's configuration.
No I am not. Users who require accessibility have it setup in their MacOS/Windows settings. It's a global setting for them. It's not website specific.
Users may access the site through a third-party device that they cannot configure, for example in a kiosk setting.
I mean we are talking about a demo project on hosting a tiny site on a calculator. Don't think anyone is trying to access it from a kiosk.

But if your comment is a general opinion of websites in general, then yea, I see what you mean.

No, it will not override the CSS (not sure what you mean by "accessibility enabled" exactly).

> even with the focus outline disabled on other sites, I am able to navigate with the keyboard to text input fields

Correct, but for a user with vision problems, the focus outline lets them see what has focus. CSS doesn't stop them from inputting values.

Focus styles are essential and should not be removed as a best practice (in general, this is a very small site).

Still hanging in there so far! Bit slow, but still faster than cnn.com for me.
Slow load, but after a minute it worked for me. Pretty cool!
Yeah. I assumed it was hugged to death. But it does load. Albeit slowly.
One of the few hn times when I try and load it still actually works, impressive little thing!
Did this get the hug of death?
It looks like it may have briefly - but it loaded great for me.
Loaded for me!

(Meanwhile, I got a "Can't serve requests that fast" from HN when I first tried to reply.)

Looks like it's time to upgrade to a Casio server!
All startups soon learn the importance of running server grade casios in production.
The calculator likely exploded
The little engine that could(n't)
based on the hostname, although the httpd isn't answering, it makes me wonder if they also made the calculator speak BGP and take a full ipv4/ipv6 BGP table. Somehow with the RAM requirements I doubt it.
lol! The IP stack does not have forwarding support, but I guess you could implement some rudimentary loopback announcement support.
Is this Casio colocated in a Hetzner datacenter?
maybe he has reverse proxy to house because no public ip with ports for hosting?
I would argue it's not really hosted on the Casio then, since it's only static content the reverse proxy cache would hit 100% every time, except on TTL.
Maybe it's just a forwarding setup with no caching?
It is cheaper than AWS's Casio as a Service and for small sites you barely notice the difference.
Then again AWS's Casio as a Service is available in a HIPAA compliant form.

Actually, HIPAA auditors would probably love a webserver running on a Casio since the hardware can be closely vetted and won't randomly change.

If you use an an alt-CaaS you'll won't benefit from Infinidash integration, but then again you won't have the Infinidash configuration learning curve to deal with so it could still end up a net positive.
Sure, but with fixed hardware resources you risk the site going down with traffic spikes like this, with AWS you have access to calculatorless options that can easily scale to handle increased request volume.
Is that their competing product offering to match virtualizaled calculators delivered by hyperdivisor?
> AWS's Casio as a Service

AWS Cassiopeia

Assuming Casios can run links or some custom browser - does anyone want to grab a snapshot of this page being rendered on their calculator to complete the circle?
According to timestamps of comments here I came approx 12 minutes late. RIP Casio
this is like firing the death star at a plucky little pod racer
"I felt a great disturbance in the Force, as if millions of voices suddenly cried out in curiosity and one was suddenly silenced."
Doesn't this run on one of the CPUs Sega used in the Saturn or Dreamcast?
Not quite the same CPU but they're all based on the SuperH ISA. The 32X and Saturn both had dual SH-2s, Dreamcast had an SH-4, and Casio uses SH-3 and SH-4. The Dreamcast SH-4 has an FPU but the Casio one doesn't.
Same as in the Dreamcast yes, albeit the Casio seems to be clocked far lower (29mhz in Casio vs 200mhz for DC). Also the memory seems a bit anemic at 60ish-kb vs 16mb main for the DC (+8mb videomem and 2mb audiomem).

It was a fine CPU for the era with parallel instructions but also a fairly deep pipe compared to other contemporaries so it did best with a bit of hand tuning. (We worked on a DC game and Transform&Lighting loops needed hand optimizations iirc)

It's not any more. Please update the title.
The author's blog has a lot of interesting projects, though I did not find a writeup on this one. https://tbspace.de/
This calculator got me through college. I then passed it down to my brother and it lasted him through college also.
Time for your brother to consider setting up a webserver on it! /s

The TI 80 series were more popular in my college days, perhaps these Casio's were more popular in another time or region?

I think Casio calculators are quite popular in the UK - I used an fx570 through secondary school and university. It was the one recommended by my school, and then it was one of the three models (all Casio) allowed by my university[1]. I briefly had a TI83 for A-Level Further Maths, but I didn't use it much.

The fx-9750 would have been usable at secondary school (though it's more powerful than what you really need), but it wasn't on my university's allowed list.

[1] Currently you can use an fx 991, fx 115, or fx 570 for CS, and I pretty sure the list was the same whilst I was there - https://www.cst.cam.ac.uk/teaching/exams/calculators

Well Casios are better than TI calculators.

My current favorite calculator is Casio ES-115 PLUS2

https://www.amazon.com/Casio-fx-115ESPLS2-Advanced-Scientifi...

I like it because when you enter sin(pi/12) you get (sqrt(3)-1)/(2*sqrt(2)) in mathematical notation, which is really nice for a $16 calculator.

Here I'm going to hurt TI: If you need a TI-84 or TI-89 for school, the place to get them cheap is shopgoodwill.com:

https://www.shopgoodwill.com/Listings?st=ti-84%20plus&sg=&c=...

Cleaned url: https://www.amazon.com/Casio-fx-115ESPLS2-Advanced-Scientifi...

Thank you so much for pointing out fx-115ES PLUS2! I looked for a good hour yesterday trying to find an alternative to the fx-115ES PLUS (first edition) with the same features, after Casio discontinued the original. I'm honestly embarrassed that I had overlooked this...

I do absolutely agree that Casio calculators are just wholly better than TI in general. Their menu driven interface is a game changer for me back in middle school. TI-36X Pro felt cumbersome and confusing, while fx-115ES was much more intuitive.

Better as it is solar… can it use similar firmware with similar c and web server?
They're amazing little machines. I have a predecessor to that model, whose battery still works since purchased new for my school year starting in 2000. It's not solar-charged, either.
I really don't understand why Casio keeps changing the entire design of their calculators, though.
they don't; they have multiple lines which don't change much generation to generation.

a casual observer might think that they can't pick a design.

I’m not sure I understand. The calculator linked is the second version of that model, which to my understanding is identical software wise but looks completely different.
TI is pretty clearly just coasting on being the one mandated by a lot of schools; there's no way you can justify them still costing like $100 and having the same capabilities as when I was in high school.
The new ones have rechargeable batteries, color screens, USB, and Python. So, only a decade out of date.
TI still use really outdated hardware (some ARM9 thing) on their high-end nSpire models.

With the same price even HP Prime G2 has way better hardware (NXP iMX6ULL, Cortex-A7, 256MB DDR3, USB EHCI OTG) and wide software spectrum (no Secure Boot enforced.) I've put full scale Linux and Windows ARM (arm32 IoT) on that platform.

Fun Fact: New Arm 9 chips are still coming out to this day. They are the highest performing Arm chips that feature generic external bus interfaces.
I searched for graphing calculator apps on iOS the other day and found a ton that simply attempt to recreate the TI-84, e.g. [1].

So anyone with high school kids: are they still buying $100 physical calculators? Or are they just using free/cheap apps?

[1] https://apps.apple.com/us/app/graphing-calculator-x84/id1247...

I'm a few years out of high school, but during my time anything other than physical calculators were explicitly banned due to the potential for kids to be accessing the internet during exams. I even had a friend who won one of TI's higher end calculators at a math competition, only for him to be banned to use it during exams because of its higher capabilities. Perhaps things have changed in the post-Covid world.
In my high school physics and math classes, we were allowed to run any programs that we programmed ourselves. This was on Ti-84s so BASIC not Python.

Given the context, it made a lot of sense.

Yeah about the same here. I wrote some programs to handle quadratic formulas. Impressed the Algebra teacher who then let me take AP Computer Science so I can learn to program in Java...which I promptly forgot after graduating HS.
When I was in HS, at the exams we were expected to use our TI-84+ calculators, but we were not allowed to have any programs of our own, so they made us clear the programs before they handed out the problems.
Which you could work around by archiving and then unarchiving the programs. I remember having to do a soft factory reset for one class before exams, rather than just clearing programs. At least they took the policy to its logical extension
Just graduated secondary school in April. Just as a someone said in a reply below, anything other than a physical calculator is still banned during assessments. There are tools like Desmos, GeoGebra, etc... that may be used by the instructor and students during lessons though. Although it's strange, there is no explicit rule provided by our Mathematics faculty that bans higher functioning calculators I've only seen one person carrying one.
That's what I did 10 years ago when my TI-89 broke right before my Linear Algebra final. I somehow convinced my instructor that when I had my phone out, I was actually using the calculator on my phone.
Oppositely here in Japan (where Casio locates), high school students usually don't use calculators on math/physics classes.
That's interesting and somewhat surprising to me. Do they use slide rules and/or function value tables printed on paper?
Not sure about Japan but in Poland the variables in problems are either symbolic and you solve it symbolically, or if they are given as values then they are usually chosen just right to not make arithmetic much of a bother (that actually results in a fun 'sanity check' that if you start getting something waaay crazier than 'normal' you double check if you didn't screw something up somewhere along the way or if there is a better way to solve it).

But everyone is taught to only substitute variables for values at the very, very end so calculators are not needed at all. At some point in education there is no point in testing arithmetic skills over and over again, the problems and the methods of finding out solutions to them are what matters

Exactly, same in the UK. So a typical first question in the maths exam we take at age 14/15 might be ‘Factorise the quadratic x²+5x+6’. No calculator needed.
Almost same as Japan. If we need calculation (like for some physics exam), we use column method on paper.
That makes sense. If you used a calculator's computer algebra system (CAS) for the symbolic manipulation, it would certainly count as cheating.
We were allowed to use calculators when I took my school exams in the uk. But our exam questions often asked us to give the answers in surds i.e 3/4*√2 not as a decimal which our calculators couldn’t do then. Our teachers told us there would never be a question where the answer couldn’t be expressed cleanly using either fractions or surds, that we should practice doing the maths and show our workings on the exam paper and that if we were using a calculator we were wasting time doing data entry as the actual addition / division / subtraction would never be hard enough to justify it. We memorised a table of surds for 30º 45º and 60º angles but I’ve forgotten why now. Also they hard reset our calculators as we went into the exam.
>Well Casios are better than TI calculators.

Agreed. For basic maths, the best scientific calculator is the fx83. The ES 115 is based on it. The graphics and controls and navigation is really good. Nothing comes close.

"This website is hosted on a Casio..."

Not anymore it isn't.

I really enjoy the HP calculators. RPN is much closer to how my brain perceives the equations.
Unfortunately they seem to have abandoned it on their newer models: I recently got a HP Prime and no RPN in sight afaict (https://www.hp.com/us-en/campaigns/prime-graphing-calculator...) .
Sure there is. Go to Settings and move down to Entry. Hit Enter and you get your choice of Textbook, Algebraic, or RPN. Go down to RPN, press Enter and then Esc.

Presto, RPN.

Thank you - I used the calculator more in the last hour after making this change than I had in the past month. I'm ashamed I didn't figure this out - I should have done a better job of RTM.
Checking in with a TI-30X Pro and get the equivalent (sqrt(6)-sqrt(2))/4 also in mathematical notation. All of my work is saved when I turn the calculator back on. Plus I can just set a flag for imaginary mode without a separate complex app so everything works the same in both modes. The screen is higher resolution and the multi-tap entry is pretty darn sweet.
Annoyingly TI-30X Pro is Europe only (also costs more, but looks better quality and design)... but you are right, it looks pretty good, as does the US-available TI-36X Pro. I will give one a try. My main gripe with TI is the school requirement for the TI-84 series.
Must be a market specific model, because that one shows up for me in the .com tld but not in the regional one :-(

But from there I've started digging and they have so many different models with similar appearances... Anyone knows the difference between fx-82ES and fx-82SPXII?

> Well Casios are better than TI calculators.

For scientific calculators, I would agree. For graphing calculators they might be better on technical merit or cost, but everyone is going to write programs for the TI so you might as well get that one.

Not everyone, TI seems to be US thing due to the government agreements.

I never owned one, and got three generations of scientific programmable CASIO calculators.

It was either that or HP ones, as TI were pretty much ignored in Portugal.

I don't know what TI ever did to you, but I know that I would not cross you lightly.
They're exploiting an effective monopoly position (through schools mandating their usage). I can see that being enough of a reason to push back against that.
Correction: was hosted on an Casio fx-9750GII Calculator.
Next post: how I burned my Casio calculator.
The author seems to have their own AS for "experimentation purposes" https://tbspace.de/as203478tbspacenetworks.html

https://bgp.he.net/AS203478

https://www.peeringdb.com/net/10981

I've always been interested in doing something like this, although it's prohibitively expensive for me, even for IPv6-only.

It can be relatively inexpensive, depending on which RIR you’re using. There are providers like [Neptune Networks](https://neptunenetworks.org/) or [Vultr](https://www.vultr.com/) that you can peer with from a VPS so you don’t need to get “proper” IP transit in a datacenter.
Oh, interesting -- I assumed you had to have a membership to the RIR directly (APNIC in my case, which would have been upwards of 1000AUD/month). Hmm... Thanks!
APNICs info seems to show AUD 1180/year, not per month? Includes one ASN, one /24 on v4 and a /48 on v6. It's still significant, but not crazy.
Oh, you're right. 1180AUD / year + 500AUD sign-up fee.. Still a bit expensive. If I had a business use case for it I'd be able to justify it, but it's too much for a hobby, unfortunately.
This actually doesn't seem that bad for a hobby (I'd be using this almost as a fee to help me learn more about networking).

Does APNIC give out ASes for hobbyist and experimental use?

You "just" need to be in their reigon, and have justification. Usual justification is that you have two upstream networks and they'll announce your prefixes (generally with BGP). That's going to cost some more money, but there's specialist VPSes for this these days.
There’s no hobbyist program but neither is it corporate/for-profit only, from what I see.

There are few people(mostly network engineers or researchers by trade and/or business owners themselves) who are assigned ASN to their homelab gears from JPNIC. Knowing the right doors to knock etc.

i got my V6 and V4 space, along with my ASN though Hostus.us (https://my.hostus.us/cart.php?gid=55). I use Vultr and a few other providers. For V6 only, you can get the ASN and V6 space for around $50 per year, and the vultr server is around $5 per month... my own ASN is under https://as204994.net with links to some providers and details...
I never would have come across this myself when doing very cursory research. One step closer... Thanks!
Excuse my ignorance, but how do you peer on public ixps through Vultr? Do you have physical presence there?
I walked through doing this with my AS (AS398328) in the US (ARIN), both for fun and for security research. Happy to chat with anyone who wants to go through the process, it was probably $1-2k in total for ARIN fees to get the AS and v4/v6 space. Real transit is a lot worse though.
1-2k is a bit tough for testing the water. But above there is another $50 one. One wonder what is the difference as $50 is very manageable!
going direct to ARIN/RIPE/APNIC etc have some stricter rules (IIRC, in RIPE you need a company or possibly business, plus a bank account in that name). by going for somone like HostUs, they are sponsoring the account, so they need to verify you and ask questions, then setup the ASN, etc. Also, going direct will provide you with V4/V6 space allocations. They might charge a yearly fee (IIRC, its like 50 per allocation) but v6 is "practally" free but V4 will be a in a wait list... Finally, going direct, there is some sort of "kick back"... if the RIR makes more money than it cost to run the place, they reallocate fees back to the LIRs... so, you some times get money back... I have been tempted, but for RIPE, its 1400 per year, plus 2k setup... ouch...

[EDIT] bit more on the RIPE fees: https://www.ripe.net/participate/member-support/payment/ripe...

I have my personal LIR with RIPE and they do allow signing up as an LIR without being a company. As an individual you need to provide proof of identity via government issued ID and pay 17% Dutch VAT.

When I signed up I got a /22 of V4 space and /29 of v6 space, although for new signups I believe there's now a wait list to get a /24 and that's it for V4.

Yeah, AFAIK the LIR model is specific to RIPE. The ARIN team does seem to put a decent amount of work into vetting and support/etc with pleasant interactions, so I don't think the fees are too unreasonable (but maybe not the best idea as an individual).
Check out the dn42 community if you simply want to play with the tech stack for free. The network uses the private AS range, allocates addresses from the private address ranges and peering happens over point-to-point VPN links.

See https://dn42.eu

Looks like the Casio is not web scale :)