My bank started using Quickbooks file format if I want to download a transaction history in a specific date range, what a fucking nightmare. It’s not abandoned yet but nothing except the QuickBooks proprietary software seems to open them so far, only a matter of time. Honestly at this point I might prefer the nightmarish CSV filetype.
The problem is that it’s not really a standard. It’s reinvented ad-hoc by whomever programs it today.
Should there be any whitespace after the comma? Do you want to use pipes or some other character instead of commas (ASCII 0x1E is sitting over there for exactly this purpose, but it’s been ignored for decades)? How do you handle escaping your separator char inside the dataset? Are you [CR] or [LF} or [CR] [LF]? None of these questions have a set answer. Even JSON has more specification than this.
Regarding “the company made the new tech incompatible with the new tech to force people to buy the new”, I’ll invoke Hanlon’s razor.
I worked for a software company that was bought out by a microscope company, because they realized making a new software from scratch for each microscope was very expensive.
They did not have the know-how to reuse the software.
And yes. They were that bad at software, when they bought us out, colleagues of mine audited the software they were writing for their newest microscope, and it was so bad they threw out the whole thing to start from scratch, with proper software engineering practices.
Also, there is an open source toolkit that is pretty good at reading microscope data called VTK (IIRC it’s developed partly by Zeiss, one of the two main microscope manufacturers).
Not just science, factory equipment that needs ancient computers to function too. If you’ve ever wondered why some old PC parts are surprisingly expensive on eBay…
Out of curiosity, I ran through some sample quizzes of the A+ exam a while back. Managed to pass, but I had to dig out a lot of my old knowledge about IDE master/slave setups and COM port settings and the like. That may be partially due to A+ being a silly, meaningless cert, but it’s pretty clear there is a need for that crap still.
I don’t know how we can’t legislate this into existence eventually if nothing else just based on climate change and the amount of working material we just… throw away. Especially as more and more things integrate software, I imagine that it’s going to feel absolutely insane to people in a few decades (after the water wars and the great migrations) that they had technology like the microscope in the post but the company decided no more software updates so now it’s just garbage.
The US (and all of their allies) are in favor of wealth redistribution to the ultra-wealthy, so predatory practices like this will never be stopped unless there is a sufficiently organized and pissed off mass movement calling for it. Researchers are a tiny group, so it won’t happen.
because “we” don’t own our government, the parasites who profit from the thing you want to change have all the power. labor needs to organize, the alternatives are capitalists killing us all or the-doohickey
The problem is that you’d need the quarter-million dollar electron microscope to test your reverse-engineered modern version, and if you get something wrong and you fry it…
That being said, I wonder why labs don’t just make a VM. Hardware passthru is definitely a thing, parallel port cards exist (as do serial port) and you can back up a VM to whatever modern storage you want. Maybe the problem is proprietary cards/connectors? PCI-X or older?
I had to reverse-engineer a floppy disk encryption scheme that was performed by some DOS software that directly talked to the IDE floppy controller. There’s no emulating that. A USB floppy drive can’t even be operated in the same way.
It was easier to just crack the (admittedly trivial) encryption.
You can get PCIe to PCI cards. I think PCIe is pretty much backwards compatible with PCI and only little logic is required. And PCI-X cards do work in PCI slots at reduced bandwidth.
Tho, if a system works without issue, why touch it? Only if parts become hard or expensive to come by a replacement makes sense.
I'm sure some do, but there's also a certain simplicity to "back up the Win95 machine" and "collect working Pentium 2's from eBay," particularly for fields that are not interested in IT for its own sake. A virtual machine adds an extra layer of abstraction and complexity, though I'm sure there's a slow trickle as entities have trouble replacing hardware or luck into technically savvy and ambitious staff. I've certainly seen my share of data being entered into a Windows 10 app that sure as shit seems to be a terminal emulator running some green-text dinosaur, or else it's got a set of Visual Basic widgets that seem like they'd be compatible with one.
The rant in the post has some merit to it, but the thing it sort of misses is also the reason not to use VM. It works just fine. It hasn’t been updated in 20 years because it still works. It does what it says on the box. Why put it in a VM? What would you gain from it? If you need Internet just grab a laptop and have it sit next to the main computer. That way users have a much smaller chance to break something vital. Pretty much all the control computers are air gapped anyway. No updates or anything to break things you reeeeally don’t want to break.
The only case I’ve seen VMs being used is if the old computer breakes and you can’t really find something that’s compatible with old-as-fuck software om bare metal. I work in a cleanroom and we got sooo many systems that are windows 95 or older (DOS anyone?). Electron microscope, etching systems, probe stations
The merit is security, as you can manage what goes into the VM as oppose to having the hardware where people can just plugin a flash drive or network cable.
Then there is also the improvement to not needing to maintain the old hardware, and having a backup of the entire system that you can just copy to a different system and have everything running again.
Sure I can see it being a security feature, random USBs are not a good thing, but I feel like it is quite minor with an air gapped system, no?
The backup is a good point. Though from this I started wondering how difficult it is to get the VM to communicate with old hardware. Like, the hardware might use some random method of actually communicating with the computer, ans getting that through to the VM might be problematic? I have no clue, just spitballing here.
Kinda related, in the company I used to work everything was done in SAS, an statistical analysis software (SAS duh) that fucking sucks. It’s used to be great, but once your on their environment you are trapped for fucking forever. I hated it and refuse to learned it over what was basic for my daily tasks. A couple of months I moved to another company that used to pay a consulting firm for my job, so my boss and me had to start everything fresh and the first thing we did was to study what are going to use as statistics software and I fight tooth and nails for Python and one of the points I pushed was that if in the future we decide to move out of Python we could easily can do it, while other solutions could locked up us with them.
If you rely on free packages in Python for processing, those are as likely to become obsolete as anything else (if not more likely). I also really dislike the compatibility issues with different versions of different packages, the whole environment aspect. Buying new computer with different version of windows? Who knows what will work there.
In this sense for scientific computation I prefer something like MATLAB. Code written 40 years ago, most likely would still work. New computer? No problem, no configuration, just install Matlab, and it runs! Yes, it costs money, but you get what you paid for. Mathematica is another option, but I mean ugh!
I mostly use pandas that I don’t think is going anywhere, we’re also going to start tests with a library called ‘chainladder’ that is used for some actuarial reserves calculations, from everything else I’m programming custom functions because as far as I know, there’s not a lot of actuarial mathematics libraries on Python (R have much more support for that, but I prefer the flexibility of Python, like a good portion of my job is scrapping our regulatory body website for information and not sure how good R work on that).
Matlab is ugly because it’s so backwards compatible. And it only is backwards compatible until someone decides to use it to interface with external hardware that you need a specific version of some library for.
If you really don’t want to spend money, there’s always GNU Octave. Sure, it doesn’t have the thousands of matlab toolboxes, but if you’re running code from 40 years ago it shouldn’t need those anyway. I wrote a couple of scripts recently and then rewrote them slightly so that they would be compatible with octave.
I work for a company who’s main source of income is a suite of accounting, stock and job management applications, all of which are written in FoxPro. The community add-ons and support is incredible but there hasn’t been any official support since like 2009.
Microsoft bought the license for FoxPro, supported it for a few years then killed it off when VB came out. I wonder why xD
The crazy part is some of our clients are turning over 100s of millions in profit a year, using this crappy, mess of a system written in a dead language, by one dude 😂
On that Windows 95 anecdote, by the way, beyond gaming that’s also one of the advantages of wine. Pretty sure their software would run perfectly on Linux with wine.
Not usually. The main thing for lab equipment is that it is controlling hardware. So you are often using proprietary drivers for custom hardware. Wine can’t handle drivers and for security reasons can’t get low level hardware access.
Not really I’m afraid. ReactOS is focused on implementing the old APIs which allows software that used those APIs to function.
But my experience with old proprietary software from that era is that they were trying to do something which was hard at the time. So they wrote their software with lots of clever tricks and hacks, to make the thing do the thing. This can be as simple as manually invoking interrupts and using undocumented APIs, to setting up non protected memory and communicating with the hardware directly.
I’ve seen cases where the software would only run with a specific version of Windows 95 and only with specific chipsets. Even changing the cpu from an Intel to a Cyrix for example could cause issues.
I was involved as an intern with a project to fix something like this a long time ago. We chose to simply reverse engineer the hardware interface, put in a custom controller to handle it and write modern software. It took a lot of doing and we lost some features, but the original system was beyond saving.
As time went on more and more hardware abstraction is applied, so I would hope this issue would be fixed in the future. But the whole of the 90s and early 2000s is a big issue.
Ah yeah, drivers are another thing entirely. Especially for what I imagine is very proprietary undocumented hardware. The only thing that can help there is a reverse engineer / kernel module dev.
When Windows dropped support for XP, our NMR lab decided to change the OS of the PC linked to the NMR machine to Linux. Since I don’t work there anymore I don’t know if they were able to do that successfully.
Not only does most scientific instrument software become abandonware, but there are companies that sell instruments that use the exact same components as they did 20 years ago. The only difference is now they swapped the stainless steel parts for plastic and charge luxury car prices for what will be a piece of garbage in 3 years. These pieces have nothing to do with chemical compatibility and everything to do with increasing the frequency of maintenance that the older models never needed.
Add comment