Skip to content
 
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Wired)   Software developers make great tools for themselves. Scientists get FORTAN, doctors and other clinicians get MUMPS. One guess where the unicorns ain't   (wired.com) divider line
    More: Sick, Climate, Climate change, Climate model, Global climate model, climate models, climate change folks, Computer program, Good software  
•       •       •

1048 clicks; posted to Business » on 24 Sep 2020 at 2:46 PM (4 weeks ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook



49 Comments     (+0 »)
 
View Voting Results: Smartest and Funniest
 
2020-09-24 2:11:26 PM  
Good software can get quite expensive. Your hospital went with the cheapest solution. It also helps having users involved in that process to lessen the chance of buying a pig.

And, there's nothing wrong with FORTAN if you're just doing a lot of math.
 
2020-09-24 2:25:50 PM  
I'm having flashbacks of being forced to learn IDL despite it being obsolete for 20 years.
 
2020-09-24 2:52:25 PM  
When I go to the doctor, they ask what I do, and when I tell them, they start complaining to me about the software at the hospital.

Sure they do.  When I do that, most people don't give a fark and don't bring up their shiatty software unless I comment on it directly.

In an IDE you drag buttons around to make the scaffolding for your apps.

...If you don't know how to XAML.  "But Dave, you can drag controls in a xaml designer!"  Yes but two of the biggest problems with that are it totally farking you xaml formatting which you *will* have to dive into eventually and that the dragging of a control auto-loads your name space +  related library which can result in all kinds of farkery (could be turn-offable...I haven't checked).  Plus, it's just "better" to write out the xaml.

In any case, doing software for scientists, engineers or similar, unless it is something tiny, is no simple task.  It often takes a developer who understands the process and can look at their database (raw data) and understand what they are looking at because chances are, even the end user themselves will not know how it all fits together and you'll be *lucky* to find a resident IT person who knows.

Why? Because they will *never* give you complete requirements.  They will always miss something very important or totally fark it up or simply not have the time or patience for building requirements.  And that's the thing - they don't or can't do such things even when they devote resources to those ends.  For my users, half the time, I'm the one leading them through their requirement writing (man, do I have stories).
/I could write a book here so don't shiat on what I'm not covering...too much..
 
2020-09-24 3:01:27 PM  
FORTRAN
 
ZAZ [TotalFark]
2020-09-24 3:05:02 PM  
You type a few letters and the software guides your hand and finishes your thoughts

I hate that sort of user interface. I first saw it around 1990. Somebody had a new shell or terminal emulator that automatically finished your commands. I was fighting the computer all the time instead of using it as a tool. I much prefer making an explicit request to complete.
 
2020-09-24 3:06:38 PM  

Tr0mBoNe: Good software can get quite expensive. Your hospital went with the cheapest solution. It also helps having users involved in that process to lessen the chance of buying a pig.

And, there's nothing wrong with FORTAN if you're just doing a lot of math.


"I don't know what the language used for science and math in the next decade will be, but it will be called FORTRAN." 1960-2010.

I think it is being replaced by Python, although the 2.x-3.x shift might complicate things.  Pypy (fast python) is still on 2.x, and getting numpy faster than it is getting 3.x (if it is bothering at all).  I don't think there are any advantages to 3.x for numerical users, and 2.x's easy handiling of strings made out of bytes is probably a significant advantage.

Something else might come along and dethrone python thanks to this division.

Does MUMPS help prevent the horror of Oracle?  I don't have the time to get around the paywall, but I suspect that it should save them.  Last I heard, MUMPS was considered amazing in the 80s-90s (probably pre-SQL or something), and is simply the dinosaur that won't die now.  But you *really* don't want to invite Oracle in your house.
 
2020-09-24 3:13:42 PM  
Yeah, you ever try writing software for a Doctor?

You'll regret it, I promise.

/if the scope creep won't get you, the endless undermining will
 
2020-09-24 3:33:16 PM  

yet_another_wumpus: Tr0mBoNe: Good software can get quite expensive. Your hospital went with the cheapest solution. It also helps having users involved in that process to lessen the chance of buying a pig.

And, there's nothing wrong with FORTAN if you're just doing a lot of math.

"I don't know what the language used for science and math in the next decade will be, but it will be called FORTRAN." 1960-2010.

I think it is being replaced by Python, although the 2.x-3.x shift might complicate things.  Pypy (fast python) is still on 2.x, and getting numpy faster than it is getting 3.x (if it is bothering at all).  I don't think there are any advantages to 3.x for numerical users, and 2.x's easy handiling of strings made out of bytes is probably a significant advantage.

Something else might come along and dethrone python thanks to this division.

Does MUMPS help prevent the horror of Oracle?  I don't have the time to get around the paywall, but I suspect that it should save them.  Last I heard, MUMPS was considered amazing in the 80s-90s (probably pre-SQL or something), and is simply the dinosaur that won't die now.  But you *really* don't want to invite Oracle in your house.


It's called M now.

You can execute SQL queries if you are so inclined.
 
2020-09-24 4:09:33 PM  
Le sigh...

The government and hospitals have spent untold billions on medical software.  Back when ICD 10 was being phased in, mid sized hospitals were being fleeced for $50-70 million each (plus annual support) for software upgrades/conversions by the various vendors (Epic, Cerner, Siemens, and a handful of others (not you McKesson)).

The problem is that the vendors have great interoperability between their own systems, but they want to keep others out of their data.   Epic would rather stick a hot old school lightbulb up it's butt than talk with Cerner.  Meanwhile the feds walk around with their thumbs up their butt pushing Direct, which is basically encrypted/secure email, but it only allows addressing to the facility level, but not individual departments or physicians.

There have been a handful of attempts to put third party systems in between them, but the systems all have different ID numbers, you can't use SSNs, and each time something is sent to the wrong place it's a $10,000 HIPAA violation.

There's FHIR out there, but it came along around the same time that I threw my hands up and decided to do something useful with my career.

yet_another_wumpus: But you *really* don't want to invite Oracle in your house.

I would use Access, SQL Lite, or text files as a database before I signed a contract with Oracle.
 
2020-09-24 4:31:12 PM  
Developers in general the laziest SOB's in the world, because we CAN actually automate our own jobs.

Can you imagine Harry Potter and company living productive, charitable lives to the betterment of their fellow man if they could just spend a few minutes saying "Accepto, farkallthisshiataco" and be done?

And yes, we will spend hours automating a 10 minute process, because that is ten minutes we don't have to spend EACH DAY.
 
Juc
2020-09-24 4:32:41 PM  
I make tools to make my life easier for free, if you want better tools, pay for 'em.
 
2020-09-24 4:40:40 PM  
FTA: "This is one of the weirdest puzzles of this industry "


No, it isn't a puzzle at all. Everything is based on the transfer of information. CS people both pay more attention to new tech and have the skillset to more rapidly adopt it. This compounds.... a lot.

Engineers got taught in MATLAB or fortran if they are old. There's a subset learning the newer stuff but that doesn't mean there is a critical mass of knowledge on newer languages in those industries.

Don't take this as me badmouthing engineers or other fields though. When i said skillset - i meant it. If CS people had more alternative domain knowledge they'd have less knowledge about programming languages. Anyone that has programming as a secondary skill is going to be less able to keep up with newer languages.

Finally, people underrate the hell out of how valuable legacy code is. Porting stuff to a new language can be one hell of a time consuming process and building something from scratch is a hell of a lot nicer on paper than in reality. If all of the engineering models are in Fortran, then people are going to keep using Fortran and that is going to self-reinforce until they want features Fortran doesn't have or run out of a labor pool that knows it.

Tr0mBoNe: Good software can get quite expensive. Your hospital went with the cheapest solution. It also helps having users involved in that process to lessen the chance of buying a pig.

And, there's nothing wrong with FORTAN if you're just doing a lot of math.



There's plenty wrong with FORTRAN. Just because it isn't as bad as it used to be doesn't make it good. Language ergonomics, compiler helpfulness and memory safety are all nice to have.

yet_another_wumpus: Tr0mBoNe: Good software can get quite expensive. Your hospital went with the cheapest solution. It also helps having users involved in that process to lessen the chance of buying a pig.

And, there's nothing wrong with FORTAN if you're just doing a lot of math.

"I don't know what the language used for science and math in the next decade will be, but it will be called FORTRAN." 1960-2010.

I think it is being replaced by Python, although the 2.x-3.x shift might complicate things.  Pypy (fast python) is still on 2.x, and getting numpy faster than it is getting 3.x (if it is bothering at all).  I don't think there are any advantages to 3.x for numerical users, and 2.x's easy handiling of strings made out of bytes is probably a significant advantage.

Something else might come along and dethrone python thanks to this division.


Did not realize Pypy is still on python 2. That is a shame. Afaik most of the momentum in science stuff is towards 3.

I wouldn't complain if Rust found a niche there though. Solid language and able to fill a lot of needs of the Fortran/C/C++ that are used today for performance.
 
2020-09-24 4:43:12 PM  

maxheck: Developers in general the laziest SOB's in the world, because we CAN actually automate our own jobs.

Can you imagine Harry Potter and company living productive, charitable lives to the betterment of their fellow man if they could just spend a few minutes saying "Accepto, farkallthisshiataco" and be done?

And yes, we will spend hours automating a 10 minute process, because that is ten minutes we don't have to spend EACH DAY.

 
2020-09-24 4:44:55 PM  
FORTAN?
 
2020-09-24 4:48:21 PM  

OptimisticCynicism: There's plenty wrong with FORTRAN. Just because it isn't as bad as it used to be doesn't make it good. Language ergonomics, compiler helpfulness and memory safety are all nice to have.


You mean all the stuff they added in 1995?
 
2020-09-24 4:49:02 PM  

ZAZ: You type a few letters and the software guides your hand and finishes your thoughts

I hate that sort of user interface. I first saw it around 1990. Somebody had a new shell or terminal emulator that automatically finished your commands. I was fighting the computer all the time instead of using it as a tool. I much prefer making an explicit request to complete.


Make it idiot-proof, and the only people who can use it are the idiots.
 
2020-09-24 4:53:16 PM  
First thing I learned as a $50 / hour developer was that if you can save an engineer or architect 10 minutes a week, you've more than paid for yourself. And they do not mind because well, company.

My rates went up accordingly in the early 1990's.
 
2020-09-24 4:56:50 PM  

gunther_bumpass: ZAZ: You type a few letters and the software guides your hand and finishes your thoughts

I hate that sort of user interface. I first saw it around 1990. Somebody had a new shell or terminal emulator that automatically finished your commands. I was fighting the computer all the time instead of using it as a tool. I much prefer making an explicit request to complete.

Make it idiot-proof, and the only people who can use it are the idiots.


Was really pissed off when all the CS slots were taken by people who didn't want to be there. I feel for you.
 
2020-09-24 5:39:28 PM  

maxheck: Developers in general the laziest SOB's in the world, because we CAN actually automate our own jobs.

Can you imagine Harry Potter and company living productive, charitable lives to the betterment of their fellow man if they could just spend a few minutes saying "Accepto, farkallthisshiataco" and be done?

And yes, we will spend hours automating a 10 minute process, because that is ten minutes we don't have to spend EACH DAY.


Developers will fark around with all their goddamn IDEs and special snowflake packages for weeks before actually doing any goddamn work.
 
2020-09-24 6:26:19 PM  

yet_another_wumpus: "I don't know what the language used for science and math in the next decade will be, but it will be called FORTRAN." 1960-2010.

I think it is being replaced by Python, although the 2.x-3.x shift might complicate things.  Pypy (fast python) is still on 2.x, and getting numpy faster than it is getting 3.x (if it is bothering at all).  I don't think there are any advantages to 3.x for numerical users, and 2.x's easy handiling of strings made out of bytes is probably a significant advantage.

Something else might come along and dethrone python thanks to this division.

Does MUMPS help prevent the horror of Oracle?  I don't have the time to get around the paywall, but I suspect that it should save them.  Last I heard, MUMPS was considered amazing in the 80s-90s (probably pre-SQL or something), and is simply the dinosaur that won't die now.  But you *really* don't want to invite Oracle in your house.


M and SQL are about the same age.  The advantage of M is that for many, many years it dealt a lot better with sparse data sets than SQL did, and any given person's medical record consists largely of sparse data (unless you are Montgomery Burns).

As someone who has used it - it's not that bad.  It's more or less like programming in C, minus having to deal with explicit memory management, and with the abstraction that a database field is just another variable that you can read and write to, one that happens to be persistent and shared across processes.

I also think the author conflates programming languages with development environments.  While it's true that more popular languages tend to have highly featured IDEs, an IDE does not need to be tied to a specific language or languages.  You can absolutely get a modern IDE for FORTRAN or M.
 
2020-09-24 6:33:36 PM  

The Googles Do Nothing: maxheck: Developers in general the laziest SOB's in the world, because we CAN actually automate our own jobs.

Can you imagine Harry Potter and company living productive, charitable lives to the betterment of their fellow man if they could just spend a few minutes saying "Accepto, farkallthisshiataco" and be done?

And yes, we will spend hours automating a 10 minute process, because that is ten minutes we don't have to spend EACH DAY.

Developers will fark around with all their goddamn IDEs and special snowflake packages for weeks before actually doing any goddamn work.


Our evil secret remains intact, then....
 
2020-09-24 6:47:45 PM  
That was an article written by somebody who was on a strict deadline and/or paid by the word.

Here's the entire article:

Old software exists.

It's mainly used by people other than computer companies.

Computer companies tend to use new software.

The end.
 
2020-09-24 6:56:26 PM  
I would love to write good software for the medical profession, and I'd like to think I have done so. But whenever you actually TRY to do proper user analysis, you get "Just make it happen, and make it whatever I read in a magazine, like, online too.

Mind-reading is an easy recipe for nobody liking anything.
 
2020-09-24 8:13:54 PM  
Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.
 
2020-09-24 8:21:37 PM  
OptimisticCynicism:

Did not realize Pypy is still on python 2. That is a shame. Afaik most of the momentum in science stuff is towards 3.

PyPy has had a stable Python 3 compatible release for a while now. It's behind the main line, (3.6) but as 3.6 is still a PSF-supported release, most major packages in PyPI are compatible.

https://www.pypy.org/download_advance​d​.html

https://www.pypy.org/compat.html

There are some large-ish internal financial apps which may be stuck on 2.7 because the move effort for 10 million loc or more is massive. 

And there will be some old stuff that never gets moved because it's not worth the effort.

But by and large most packages and projects have been moved, and language adoption (contrary to all the doom and gloom predictions) is still growing, at least for now.

Also most scientific/math/ml/ai packages are written in native-compiled code with a Python wrapper (simplifying a bit for the non-Pythonistas). Those kinds of packages can actually be slower on PyPy.
 
2020-09-24 8:23:37 PM  

Russ1642: Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.


Sounds thoughtful. Want to elaborate on that?
 
2020-09-24 8:48:10 PM  
OptimisticCynicism:
Engineers got taught in MATLAB or fortran if they are old. There's a subset learning the newer stuff but that doesn't mean there is a critical mass of knowledge on newer languages in those industries.

Engineering students are still taught MATLAB and Simulink, which are updated twice a year. The "toolbox" model continues to have demand. Even "new" projects such as Space-X Dragon ran guidance developed with these tools.

It's probably time to update your copy of MATLAB.
 
2020-09-24 8:51:04 PM  

MusicMakeMyHeadPound: Yeah, you ever try writing software for a Doctor?

You'll regret it, I promise.

/if the scope creep won't get you, the endless undermining will


Writing research project applications for university medical PhDs was bliss.  Some of the nicest people I've ever met.

Practitioners unfortunately get thrown in the same bucket as lawyers.
 
2020-09-24 9:00:53 PM  
In the sea being held at bay by the Red Bull?
 
2020-09-24 9:05:04 PM  

maxheck: The Googles Do Nothing: maxheck: Developers in general the laziest SOB's in the world, because we CAN actually automate our own jobs.

Can you imagine Harry Potter and company living productive, charitable lives to the betterment of their fellow man if they could just spend a few minutes saying "Accepto, farkallthisshiataco" and be done?

And yes, we will spend hours automating a 10 minute process, because that is ten minutes we don't have to spend EACH DAY.

Developers will fark around with all their goddamn IDEs and special snowflake packages for weeks before actually doing any goddamn work.

Our evil secret remains intact, then....


worked in COBOL for Dept of Treasury Y2K projects.  I discovered how to program a destructive backspace* into the IDE.  I felt like a god since everyone's efficiency went way up.  $12 an hour intern at the time.

*before that you would have to cursor back one spot and then hit the delete key
 
2020-09-24 9:05:30 PM  
The problem I've encountered w/ hospitals, is that they usually have deep purses, they also have doctors who spent 8 years in school becoming the best and the brightest. Which means they don't listen to the IT people. Worse is that vendors know this and will play on the MD's ego, only offering them the best of deals, and the IT guys are only against it because they are jealous of that huge MD intellect. Etc, Etc and so on, and such.

Sounds like not much has changed.

/ 10,000 years of history and the human society has not advanced one iota.
// We are still only 12 meals away from going full on Lord of the Flies.
 
2020-09-24 9:08:05 PM  

nemisonic: OptimisticCynicism:
Engineers got taught in MATLAB or fortran if they are old. There's a subset learning the newer stuff but that doesn't mean there is a critical mass of knowledge on newer languages in those industries.

Engineering students are still taught MATLAB and Simulink, which are updated twice a year. The "toolbox" model continues to have demand. Even "new" projects such as Space-X Dragon ran guidance developed with these tools.

It's probably time to update your copy of MATLAB.


yea, I took Andrew Ng's machine learning course last fall via Coursera/Stanford.  He was in essence using MATLAB.   He worked with Octave, a free variant of Matlab., but syntax was nearly identical.
 
2020-09-24 9:15:20 PM  
ease of use isnt the same thing as efficiency.  their use case of simming temp changes of every ocean on earth vs building some little shiatbox with clicky buttons is a joke.  one of those is an insanely complicated problem of complex math that even at maximum optimization tuned specifically for solving that problem takes forever to run on a supercomputer.  the other is something you slap together in visual studio out of pre-made components in 5 minutes b/c efficiency DOESNT MATTER.
 
2020-09-24 9:51:33 PM  

Xzano: The problem I've encountered w/ hospitals, is that they usually have deep purses, they also have doctors who spent 8 years in school becoming the best and the brightest. Which means they don't listen to the IT people. Worse is that vendors know this and will play on the MD's ego, only offering them the best of deals, and the IT guys are only against it because they are jealous of that huge MD intellect. Etc, Etc and so on, and such.

Sounds like not much has changed.

/ 10,000 years of history and the human society has not advanced one iota.
// We are still only 12 meals away from going full on Lord of the Flies.


ive worked as a DBA in healthcare for 20 years. I often say that Doc's are the stupidest intelligent folks on the planet. The more prestigious the Doc, the level of stupidity grows exponentially
 
2020-09-24 10:02:50 PM  

Fano: In the sea being held at bay by the Red Bull?


Schmendrick frowns on your shenanigans.
 
2020-09-24 11:28:12 PM  

maxheck: Russ1642: Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.

Sounds thoughtful. Want to elaborate on that?


They don't want to upgrade their systems. The software is ancient. The hardware is ancient. I've read countless stories about these archaic medical databases and the people that have to maintain them. Nobody wants to spend any money to move to more modern systems. Nobody wants to put in the effort required to even develop more modern systems. And there's so much money to be made in maintaining these dinosaurs. It's a recipe for errors and failure.
 
2020-09-24 11:50:06 PM  

Russ1642: maxheck: Russ1642: Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.

Sounds thoughtful. Want to elaborate on that?

They don't want to upgrade their systems. The software is ancient. The hardware is ancient. I've read countless stories about these archaic medical databases and the people that have to maintain them. Nobody wants to spend any money to move to more modern systems. Nobody wants to put in the effort required to even develop more modern systems. And there's so much money to be made in maintaining these dinosaurs. It's a recipe for errors and failure.


Works for me. I don't disagree one bit.
 
2020-09-25 1:02:06 AM  
MDs and nurses are scared and lazy. Like, "don't want to learn a new EMR so I'll use the old shiatty one rather than adapt" and then complain that they've got to key in data in an extra dashboard. Plus good luck finding developers and maintenance for a better but nonstandard EMR.
 
2020-09-25 1:17:57 AM  

nemisonic: It's probably time to update your copy of MATLAB.


Honestly, the only thing i use matlab for is exposing a simulink interface.

MATLAB is a decent tool that costs about 10x what it should.
 
2020-09-25 1:24:44 AM  

bcostlow: There are some large-ish internal financial apps which may be stuck on 2.7 because the move effort for 10 million loc or more is massive.


Yeah, doing anything with a multi-million line legacy codebase is a royal pain. I manage one (~25 years old with over 5 mloc) and we'd need a lot more manpower to update it anywhere near as quickly as i'd like.

Guido's work at dropbox definitely helped build out public migration tools for 2->3 but it isn't perfect. As I can't control the environment clients run my stuff on, our python had to get updated to be fully 2/3 compatible.
 
2020-09-25 1:28:02 AM  

MadHatter500: OptimisticCynicism: There's plenty wrong with FORTRAN. Just because it isn't as bad as it used to be doesn't make it good. Language ergonomics, compiler helpfulness and memory safety are all nice to have.

You mean all the stuff they added in 1995?


No, I really don't. Just because something improves in an area doesn't mean it has caught up with other stuff that actually does a good job. For example, I'm willing to bet the newest fortran compilers don't offer the compiler feedback that Elm has.
 
2020-09-25 1:30:00 AM  

nemisonic: The "toolbox" model continues to have demand. Even "new" projects such as Space-X Dragon ran guidance developed with these tools.


Quite aware of this fact. My employer has dozens of MATLAB licenses of various types.
 
2020-09-25 1:30:33 AM  

Russ1642: maxheck: Russ1642: Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.

Sounds thoughtful. Want to elaborate on that?

They don't want to upgrade their systems. The software is ancient. The hardware is ancient. I've read countless stories about these archaic medical databases and the people that have to maintain them. Nobody wants to spend any money to move to more modern systems. Nobody wants to put in the effort required to even develop more modern systems. And there's so much money to be made in maintaining these dinosaurs. It's a recipe for errors and failure.


Eh, I work in the industry and I don't see that, at least not for the core EMR itself (ancillary systems are another beast entirely).  Any health system of any decent size needs to be on software that is regularly updated, if for no other reason than the rapidly-shifting regulatory requirements that CMS and ONC keep rolling out.

The biggest obstacle to developing a new EMR system is that Cerner and Epic are deeply entrenched, with four-plus decades of work behind them.  Sure, a lot of that is reworking and improving upon older functionality, but they each have such a breadth of functionality that any competitor would take on the order of tens of thousands of person-years to reach feature parity, let alone to innovate.  People aren't coming from paper these days - gone are the days when a minimal system will meet people's needs.  Today, the market is people who already have an EMR and might be willing to switch, and that means you need to offer something better.

Most of the growth areas I see in medical software are for niche apps, not EMRs.
 
2020-09-25 5:14:39 AM  
Fark user imageView Full Size
 
2020-09-25 6:40:11 AM  

Russ1642: maxheck: Russ1642: Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.

Sounds thoughtful. Want to elaborate on that?

They don't want to upgrade their systems. The software is ancient. The hardware is ancient. I've read countless stories about these archaic medical databases and the people that have to maintain them. Nobody wants to spend any money to move to more modern systems. Nobody wants to put in the effort required to even develop more modern systems. And there's so much money to be made in maintaining these dinosaurs. It's a recipe for errors and failure.


Not to mention all the patient-monitoring devices that connect via RS232, and matrix printers with parallel port connectivity...
 
2020-09-25 6:46:44 AM  

OptimisticCynicism: bcostlow:

Yeah, doing anything with a multi-million line legacy codebase is a royal pain. I manage one (~25 years old with over 5 mloc) and we'd need a lot more manpower to update it anywhere near as quickly as i'd like.

Guido's work at dropbox definitely helped build out public migration tools for 2->3 but it isn't perfect. As I can't control the environment clients run my stuff on, our python had to get updated to be fully 2/3 compatible.


Library package or a full blown app? The latter is a place where Python doesn't have a great story. Distribution of apps onto machines outside of one's control. As much as I love the language, I would not want to work on an app distributed to end users.
 
2020-09-25 7:05:43 AM  

Sum Dum Gai: Russ1642: maxheck: Russ1642: Bull. Shiat. The ONLY reason they're stuck with shiat software is their outright refusal to change.

Sounds thoughtful. Want to elaborate on that?

They don't want to upgrade their systems. The software is ancient. The hardware is ancient. I've read countless stories about these archaic medical databases and the people that have to maintain them. Nobody wants to spend any money to move to more modern systems. Nobody wants to put in the effort required to even develop more modern systems. And there's so much money to be made in maintaining these dinosaurs. It's a recipe for errors and failure.

Eh, I work in the industry and I don't see that, at least not for the core EMR itself (ancillary systems are another beast entirely).  Any health system of any decent size needs to be on software that is regularly updated, if for no other reason than the rapidly-shifting regulatory requirements that CMS and ONC keep rolling out.

The biggest obstacle to developing a new EMR system is that Cerner and Epic are deeply entrenched, with four-plus decades of work behind them.  Sure, a lot of that is reworking and improving upon older functionality, but they each have such a breadth of functionality that any competitor would take on the order of tens of thousands of person-years to reach feature parity, let alone to innovate.  People aren't coming from paper these days - gone are the days when a minimal system will meet people's needs.  Today, the market is people who already have an EMR and might be willing to switch, and that means you need to offer something better.

Most of the growth areas I see in medical software are for niche apps, not EMRs.


Regionals in rural areas can still be behind the curve.

A phenomenon I've seen is getting a shiny new Epic or Cerner EMR and saving on implementation costs by only porting records deemed active. The old system gets archived (i.e. PDFs) or just kept around.

One place I encountered had OB records (kept a long time for legal reasons) scattered across three systems. The oldest running on 1980s, pre-HP Apollo workstations. In 2016.

But it's not that they don't want to upgrade all of it, it's that they don't have the money.
 
2020-09-25 9:03:05 AM  

nemisonic: OptimisticCynicism:
Engineers got taught in MATLAB or fortran if they are old. There's a subset learning the newer stuff but that doesn't mean there is a critical mass of knowledge on newer languages in those industries.

Engineering students are still taught MATLAB and Simulink, which are updated twice a year. The "toolbox" model continues to have demand. Even "new" projects such as Space-X Dragon ran guidance developed with these tools.

It's probably time to update your copy of MATLAB.


It looks like Mathworks is porting MATLAB to gpus (googling says "just use gpuArray instead of normal arrays").  Granted, this is a bit iffy since consume GPUs are neutered to "only" work with singles and "real" numerical (non-DSP) stuff needs doubles (i.e. the expensive GPUs), but if you can convince your numerical algorithm to work on a GPU, there's no way to go faster.

I studied engineering back in the 1990s, and we used not only had to learn FORTRAN(77) but did so on the old Univac 11xx series computer (shut down right after I was done with it.  Really annoyed I had to use the thing because they didn't spring for a real editor and you had to use the univac equivalent of ed, i.e. a line editor).  I was later poking around the basement and found a Univac FORTRAN programming manual.  I was wondering why it didn't seem more familiar when I slowly realized that it wasn't mine.  It was my mom's old univac FORTRAN manual from before I was born.

/ also had to learn assembler on that for an EE computer architecture class
//no fun at all
///but x86 was probably even worse (I coded a lot of x86 assembler)
 
2020-09-26 2:59:48 AM  

bcostlow: OptimisticCynicism: bcostlow:

Yeah, doing anything with a multi-million line legacy codebase is a royal pain. I manage one (~25 years old with over 5 mloc) and we'd need a lot more manpower to update it anywhere near as quickly as i'd like.

Guido's work at dropbox definitely helped build out public migration tools for 2->3 but it isn't perfect. As I can't control the environment clients run my stuff on, our python had to get updated to be fully 2/3 compatible.

Library package or a full blown app? The latter is a place where Python doesn't have a great story. Distribution of apps onto machines outside of one's control. As much as I love the language, I would not want to work on an app distributed to end users.


So my primary code is C/C++/Fortran, but we use python for scripting stuff. It is a very idiosyncratic project, so my primary customers build their own releases of the software. That means scripts that should be internal only have to work externally.

As for deployment of Python Apps, there are a couple different avenues so it might not be as bad as you think. However, they are all pretty transparent to get source out of which would not work with my primary app.
 
Displayed 49 of 49 comments

View Voting Results: Smartest and Funniest

This thread is closed to new comments.

Continue Farking





On Twitter



  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.