What could possibly go wrong? DOGE to rapidly rebuild Social Security codebase.

A safe and proper rewrite should take years not months.

See full article...

 

[mozbo]

NO. Just no. AI is not a magic wand. Let's not call it AI but what it actually is: trained neural networks, so, machine learning. An ML system knows nothing until you train it. And you can't train it by throwing 60 MLOC of COBOL and adjacent modules at it. That achieves nothing; the resulting model still won't know how to meaningfully analyze the code.

FYI, systems for reading legacy code and extracting all its spaghetti dependencies do exist, have for a long time. But guess what -- even knowing how different parts of the code are affected by changes to other parts tells you exactly nothing about what the code does, and even less about what it's supposed to be doing. Static call graphs, dependency graphs, interaction diagrams and what have you are useful tools for the systems analyst; but they're just that, tools. And no, you can't replace the analyst with AI ... at least not quickly. Definitely not in in a few months.

It takes a bunch of people with institutional knowledge and actual experience with the system months if not years to even understand what the system does, and a bunch of other people some more years to design a replacement: that's before you write a single line of actual code (yeah, code generation might be automatable, but again, it's not at simple as it sounds). And all the while the legacy system has to keep running and keep getting changed, it's a moving target.

Only an idiot with zero experience and an overblown ego would dare to predict they can replace such a behemoth in a few months. Only a fool would consider it a good idea to try. How does the saying go? If it ain't broke, don't fix it? Yeah, it's not broken. Are there bugs, sure. Maybe work on that instead? But oh, finding and fixing bugs isn't sexy, right, that's a job for peons.

If I had an euro for every time some dingbat said, "let's rewrite it in <current fad language>" and then failed miserbly, I'd be a millionaire.

And those tools are generally useless for analyzing scaling and parallelism in gigantic systems. They will miss scaling-related changes. Not a maybe, not a probably. WILL. In systems the size of SSA, those will be lethal problems.

The only way those scaling issues get identified and addressed is by lengthy (months or years) stress testing.

Trying to do that in production - as DOGE is clear planning on - WILL cause systemic crashes.

 

[graylshaped]

I mean Elon was spot on about the timing of fully self-driving Teslas and its robotaxi fleet. I'm not sure what everyone is worried about.

<sarcasm>

It is indisputably true that FSD will and always will be working as promised Real Soon Now™.

 

[mozbo]

No, I think that honor falls on Charles Babbage. And he may have been dimly aware of this when the world's first official Luser saw fit to invoke the sort of question which through the ages has prompted old BOHFs to reach for their choice of LART.

"On two occasions I have been asked, 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question."

MIGO.

Musk In, Garbage Out.

 

[Sjoerd Verweij]

Gates knew to look for people smarter than him.

Yeah, exactly that -- you can complain all you want about Microsoft in the 1990s, but they put insanely smart people in charge of Just Making Things work, and they did. I am really good at what I do, but I will happily admit I am not (and will never will be) as good as Raymond Chen.

 

[graylshaped]

MIGO.

Musk In, Garbage Out.

To complete the syllogism: Musk, therefore garbage.

 

[alisonken1]

Yeah, exactly that -- you can complain all you want about Microsoft in the 1990s, but they put insanely smart people in charge of Just Making Things work, and they did. I am really good at what I do, but I will happily admit I am not (and will never will be) as good as Raymond Chen.

Until you got a BSOD.

Binary registry with no alternative backup? What could possibly go wrong!

 

[Sjoerd Verweij]

And those tools are generally useless for analyzing scaling and parallelism in gigantic systems. They will miss scaling-related changes. Not a maybe, not a probably. WILL. In systems the size of SSA, those will be lethal problems.

The only way those scaling issues get identified and addressed is by lengthy (months or years) stress testing.

Trying to do that in production - as DOGE is clear planning on - WILL cause systemic crashes.

Scaling is the absolute least of your worries here.

All you need to do is crank out a bunch of checks every month, right?

Welp, yeah, but that's from a master register that takes 53 semi-live feeds, so you need to keep those up and resolve discrepancies.

Then you need to accumulate/aggregate things like employment history, payment history, disability history, and thus payments due every month to single-day resolution.

All pretty straight-forward.

Now graft 90 years of exceptions and extra classifications and all kinds of cruft on top of that. I can tell you what that model looks like: it's going to have exception table after exception table and cross-referencing batch-jobs updating them for days and days and days.

If you were to give me this job, and I could bring 10 of the best I know with me, and you give me 6 months, I have no doubt we'd be able to crank out 73 million checks.

That's the easy part.

I would not in any way expect to be more than 90ish percent accurate at that point.

Give me a few more months, 95%. A year, 99%.

That's super-optimistic "I know what I am doing and I am only working with people I know and there will be NO, I mean NO TPS reporting bullshit along the way" developer estimation.

99% correct out of 73 million checks would mean 730,000 people would incorrectly get, or not get, or get a wrong version of their checks -- and that would collapse the entire SSA, not to mention get thousands of people killed from hunger or cold. I don't know why math is so hard for people. The scale of something like this is brutal. Reportedly, the saying in the Windows division was "that's a one in a billion chance!", "yes, that means next Tuesday".

 

[westernwear]

As far as the Trump administration (and especially Musk) is concerned, if people don't get paid, that is a good thing, because it saves money.

Of course, if they make a mess of it, will old people who stop receiving their checks ever vote Republican again?

Will any American ever vote again?

 

[graylshaped]

Scaling is the absolute least of your worries here.

All you need to do is crank out a bunch of checks every month, right?

Welp, yeah, but that's from a master register that takes 53 semi-live feeds, so you need to keep those up and resolve discrepancies.

Then you need to accumulate/aggregate things like employment history, payment history, disability history, and thus payments due every month to single-day resolution.

All pretty straight-forward.

Now graft 90 years of exceptions and extra classifications and all kinds of cruft on top of that. I can tell you what that model looks like: it's going to have exception table after exception table and cross-referencing batch-jobs updating them for days and days and days.

If you were to give me this job, and I could bring 10 of the best I know with me, and you give me 6 months, I have no doubt we'd be able to crank out 73 million checks.

That's the easy part.

I would not in any way expect to be more than 90ish percent accurate at that point.

Give me a few more months, 95%. A year, 99%.

That's super-optimistic "I know what I am doing and I am only working with people I know and there will be NO, I mean NO TPS reporting bullshit along the way" developer estimation.

99% correct out of 73 million checks would mean 730,000 people would incorrectly get, or not get, or get a wrong version of their checks -- and that would collapse the entire SSA, not to mention get thousands of people killed from hunger or cold. I don't know why math is so hard for people. The scale of something like this is brutal. Reportedly, the saying in the Windows division was "that's a one in a billion chance!", "yes, that means next Tuesday".

And you did the easy math.

Now compound this monthly.

 

[westernwear]

60MLOC... With a system this size, if done properly, the analyis phase, e.g. what all this code actually does under which circumstances, would take longer than a few months.
Equally, analyzing what it actually should do according to federal, state and local regulations is probably a similarily complex task.
Then writing a replacement based on those 2 inputs (and their gaps) again takes time.

The suggested timeline alone tells us, they don't plan to do this in any way you could call "properly", instead they will most probably opt for an "AI"-assisted garbage-in - garbage-out process, which might conclude in a few months with an engineering catatrophe.

What, exactly, is the reason for the rush? Why does this have to happen this calendar year? There was a migration plan that was deferred due to covid. How hard could it be to just fund and reactivate that plan? Like a smart person would do>?

 

[Sjoerd Verweij]

Binary registry with no alternative backup? What could possibly go wrong!

Just to save both of us a lot of time: are you at least remotely aware of what coding on Windows was like before 95/32s/NT?

 

[Sjoerd Verweij]

And you did the easy math.
Now compound this monthly.

"Compound"? Are you high? These are people that do not know the difference between "debt" and "deficit". I am going to explain compound interest to them? I'd rather teach origami to sea slugs, and I'd probably have the same success rate doing it.

What, exactly, is the reason for the rush? Why does this have to happen this calendar year?

Because we're being run by Business Men(TM) now, and they are going to show us how much more efficiently Business Men(TM) can run things like a Business(TM) and Get Things Done(TM).

It's feeble overpaid lazy government Checkers(TM) that say this will take at least five years; if we bring in the proper Doers(TM) it'll just get done without all the whining about grandma not getting her check.

There was a migration plan that was deferred due to covid.

Yep, 5 years (which in and of itself was kind of optimistic).

How hard could it be to just fund and reactivate that plan? Like a smart person would do>?

Because fuck you, you Checking piece of shit. We are Doers and we're going to make things happen. And yes, we're going to break all kinds of things, but we're really good at fixing it*, and no grandma would complain about her check being late, it's only fraudsters that would, so this is a great way to suss out fraud**.

* Sure, maybe half the employees have left at that point, but we can just beg them to come back for less money, but oh, wait, turns out we don't have contact info for most of them, and this is really weird, it turns out that a lot of the best and most senior employees don't want to come back at all. They must be some weird kind of Trump-hating freaks... you know, not wanting to work for an employer that fires you one day and rehires you the next, cutting all your funding, contradicting everything you do and prove.... traitorous bastards to a one.

** Genuine question; has DOGE found a single actual dime of fraud yet? Last time I checked they hadn't. This "argument" was put forth by one of Trump's billionaire leeches (gosh, I really love being told how to manage my money by a billionaire, it just gives me ALL the Luigis), I just watched it yesterday, and I can't find it. I'd love a ref.

 

[mozbo]

Legal requirements are just that, a small subset of the system requirements, which the system design has to be built to fulfill. The design does have to take them into account, because there are going to be a lot of rules you need to be able to program and manage over time.

Those couple words should've put this entire thread to rest.

 

[mozbo]

Scaling is the absolute least of your worries here.

Agreed. There's a gazillion issues with DOGE's "approach". Any number of them are system-killers. The whole thing is just a massive head-desk.

I had a co-worker, every Monday he had to review code from an offshore team of very junior devs. Like clockwork, right around 10 AM, we'd all hear his chair rotate and his head hit the desk. Someone drew a target on a piece of paper, captioned it "Your Head Here", and put it on his desk. He kept that paper for years.

 

[Sjoerd Verweij]

Agreed. There's a gazillion issues with DOGE's "approach". Any number of them are system-killers. The whole thing is just a massive head-desk.

To be clear: I'd LOVE to get this contract, guaranteed and up-front, for, IDK, 3 years? There would be absolutely 0 need to check anything in, or write anything, because this WILL be exposed as an idiotic boondoggle and killed off in 2027, 2031 at the latest. None of this has to work because it will never will work. I bet if you'd roll into this project in a month or two you'd be working off a crayon drawing by BigBalls for architecture.

Edit: I mean I could day-drink from 7AM and pass out by 10AM... why cares? None of this will ever go live or meaningfully move forward; it'll be a shitload of shit contractors writing 50MB of shit Java code and maybe a Jenkins script or two, and NOTHING will be done, and NOTHING will change. Faaahk, like we're new here guys. I've seen IBM, Tata and Globant do exactly this, live and up front, at least half a dozen times.

Edit 2: Of COURSE it can actually be done. It will take 3-5 years, and it would require the state to contract out themselves rather than try to get this done with barely literate dingles foisted upon them by garbage contracting agencies.

 

[Komarov]

So that's a no then? Because everything you've mentioned are simply implementation details that are required solely to have a functional system that implements the legal aspect.

And protip: Insulting people on the internet who you don't know isn't a good look.

Telling you that you have no clue what you're talking about is not an insult, it's a statement of fact. And it's "whom", not "who".

 

[Komarov]

What, exactly, is the reason for the rush? Why does this have to happen this calendar year? There was a migration plan that was deferred due to covid. How hard could it be to just fund and reactivate that plan? Like a smart person would do>?

Your question contains the answer.

 

[alisonken1]

Just to save both of us a lot of time: are you at least remotely aware of what coding on Windows was like before 95/32s/NT?

Actually, yes.

PC/XT clones.
286/386 computers.
Pentiums.
Windows 3.x
O/S2 Warp 4 was pretty good as well.

Also, Navy combat computers. I had source code on microfiche, so any changes I made to software were done using machine code (1 step lower than Assembler).

EDIT: Also Sun O/S and AT&T Unix on mainframes at a Navy research facility.
EDIT2: Speaking of which, at least 1 of our researchers was the equivalent of a Windows Insider.

 

[Madestjohn]

Will any American ever vote again?

I’m sure there will be plenty of Maga politicians and Manosphere influencers on Dancing with the Stars to vote for

 

[Fritzr]

FSD is going to be operational by no later than the end of this year at most (stated in 2017). Was a timeline ever provided for when the (totally failed) Hyperloop was going to be fully functional?

To be fair, HyperLoop was thrown out as a concept along with a challenge to anyone wanted to try to create a practical implementation. This was not an actual Musk project.

The result is that the HyperLoop challenge resulted in a lot of failures. The concept though lives on under a number of names. Most commonly maglev in a tunnel designs are HyperLoop variations, just as HyperLoop itself was a variation of train in a pressure controlled tunnel, a concept that is as old as subway systems.

 

[Fritzr]

Sounds like a training issue. Too bad corporations stopped training anybody for anything 30 years ago.

Training is expensive. It is much cheaper to just hire someone trained elsewhere.

Key the bitching about no one being trained for the critical task so that we can hire them afterward

 

[traumadog]

That was the soft tag. The hard facts:

$250 million for climate finance development accelerator to mobilize $2.5 billion in public and private climate investments.
$2 million for sex change operations in Guatemala
$2.5 million for electric vehicles in Vietnam
$6 million for Egyptian tourism
$20 million for irrigation canals and farming equipment in Afghanistan
$50 million in Gaza birth control

Just a few, but All totaling $20 billion and rising.

AID for the world. All of this sent To NGOs. Yeah, “ non governmental organizations “ , what could go wrong?

1) $250 million for climate finance development. And you think US Banks won't back future projects? How about other countries paying for US technology for these projects?

2) $2 million in healthcare for Guatemala. Not all of it (or even a meaningful percentage) went to sex changes. Did they finance LGBTQ+ healthcare? Sure. But there's more to that than sex changes. But I guess you think healthcare for gay folks is "waste".

3) $2.5 million for EV's in Vietnam - no, it's $2.5 million for startup projects for the local economy.

4) $6 million for Egyptian "tourism" was actually a grant to Egypt, signed for under the first Trump Administration in 2019.

5) $20 million for "farming and irrigation in Afghanistan" was actually food assistance... in 2017. USAID also spent $20 million on children's programming in Afghanistan recently, too. You know, "changing hearts and minds" type stuff.

6) "$50 million in Gaza birth control" - is more healthcare across Africa in it's entirety. There's no way the US shipped 1 billion condoms to Gaza. Never mind that the "Gaza" in question was in Mozambique, not in Palestine.


So ALL of these "wastes" put together sum to $330.5 million. Now you might think these are "big numbers", but they're not.

This is the equivalent of what Elon made in less than 15 hours in 2024.

All of it - and I mean ALL of it - could have been paid for if Elon's taxes went up by just 0.15% in 2024.

That's basically in the territory of a rounding error.


And if you don't think that the Administration's efforts aren't changing hearts and minds against America right now, we'll just know that Japan and South Korea are talking with China as to how to jointly deal with this tariff issue.

Mind you, China, Japan and South Korea jointly and all just behind the US in GDP. Add India, and they blow the US out of the water.

 

[Fritzr]

I don't know where you think Musk has displayed "extreme success" in terms of delivery. I will grant he used to have some PT Barnum flair (remember the flame throwers?).

SpaceX aside, nothing else is really that great.

Full self drive - nope
Solar roofs - nope
Cheap tunnels -.nope
Cheap grid scale storage - not at scale afaik
Hyperloop - nope
Neuralink - is about the same as competitors afaik
Teslas - unable to compete with overseas products on price or build quality
Robots - that's gone very quiet
Ai - Grok is fine, but it's not qualitatively different than the others

He is always over promising and under delivering, apart from the one time his engineers delivered a miracle (SpaceX). His trick is to start talking about the next thing before people become disillusioned with the previous thing

His takeover and redesign of the custom car builder Tesla did have a good outcome. It spurred everyone else to start building EVs because Tesla the EV builder proved that there was a viable market for EVs priced above what most consumers were willing to pay for a new car.

After that Tesla failed to continue to innovate and has never resolved their QA issues. The predictable result is that they are THE name brand in EVs and quickly becoming an also ran in that market segment as the competitors are now selling the EVs they have spent the last 3 to 5 years designing and debugging. Even the Chinese BYD is offering what is considered a quality product in spite of the well deserved Chinese reputation for putting profits ahead of quality.

Tesla might have coasted a few more years on the basis of being the #1 name while they fixed their QA problems, but their own Dear Leader has thrown his sabot in the works with his extreme attempt to show his expertise in managing the administrative systems of a sovereign nation.

HyperLoop was never a serious project. It was just a concept thrown out to anyone who wanted to try to make it work in an effort to hype Musk & Tesla Motors

SpaceX and PayPal (who threw him out before he could destroy them) are his only genuine successes.

 

[Fritzr]

Java is still being developed.

Project Panama: Interconnecting JVM and native code

As is COBOL. The latest releases support modern OO programming.

In development just means that all your code has to be regularly updated because the improvements to the language no longer support legacy code.

Most COBOL users beat that problem by locking in the version of COBOL they need and don't migrate as each new & improved version is released. COBOL84 is still alive because it simply works without the need for multi-million dollar/pound/euro investments to transition to the new & improved release. (Still cheaper than transitioning to an entirely different language that has no long term support because new & improved is always better)

 

[Madestjohn]

His takeover and redesign of the custom car builder Tesla did have a good outcome. It spurred everyone else to start building EVs because Tesla the EV builder proved that there was a viable market for EVs priced above what most consumers were willing to pay for a new car.

After that Tesla failed to continue to innovate and has never resolved their QA issues. The predictable result is that they are THE name brand in EVs and quickly becoming an also ran in that market segment as the competitors are now selling the EVs they have spent the last 3 to 5 years designing and debugging. Even the Chinese BYD is offering what is considered a quality product in spite of the well deserved Chinese reputation for putting profits ahead of quality.

Tesla might have coasted a few more years on the basis of being the #1 name while they fixed their QA problems, but their own Dear Leader has thrown his sabot in the works with his extreme attempt to show his expertise in managing the administrative systems of a sovereign nation.

HyperLoop was never a serious project. It was just a concept thrown out to anyone who wanted to try to make it work in an effort to hype Musk & Tesla Motors

SpaceX and PayPal (who threw him out before he could destroy them) are his only genuine successes.

Elon’s only part of PayPal’s success was still being on the paperwork when it was sold
and nothing that contributed to that sale came from Elon, not the name, not the software, design or business plan

that was all from the other side of the merger

 

[Fritzr]

(note: the following post discusses early 20th century stereotypes)

There was a not-too-distant time when secretaries were stereotypically male, and computers were stereotypically female. It was assumed that communications tasks and organizing the schedules and meetings and such shouldn't be left to women, who were much better with rote calculations.

When electronic computers were first invented, it made sense that human computers would be the ones punching the cards and feeding them into the electronic computer, and of course wiring it and setting it up.

The men wrote the equations of course, all solid mathematicians. Converting it into all that bizarre logic and wiring was left to the women.

Yes, really. It's why John von Neumann gets credit for writing about a theoretical computing architecture while his wife Klara was busy writing, programming, and running the Monte Carlo simulations on ENIAC that helped build the first implosion-style nuclear bomb.

It definitely didn't help that her work was classified and not made public until many years later when no one cared much about that part of the bomb project.

The white male theoreticians who wrote the specs that the engineers turned into hardware on the other hand are celebrated heroes and the fact that they led the effort wasn't classified, just the details of how it was implemented, which is the level at which Klara Von Neumann worked.

 

[Fritzr]

This. What “change” is necessary with SSI? It works. People get their checks. They eat and pay their bills. What’s the problem here?

The problem for conservatives regardless of their declared affiliation is that the recipients are eating and paying their bills using government funded UBI.

Albeit a limited form of UBI with gatekeeping and an absolute ban on any attempt to become self-sufficient. These restrictions can likely be traced to conservatives scared that SSI recipients might be able to break free from government control if given even half a chance. No one has ever accused the conservatives of thinking clearly

 

[Fritzr]

Probably. But most people are missing the reasons why.

When I was doing this stuff for my employer (a public university) I was in the very enviable position of doing implementation and policy. I was both technical and domain expert. In terms of broad technical expertise I'm maybe a 5/10. I'd hazard that most people on this forum are better programmers than I am. But I can hold my own. I've been coding for almost 45 years - there's experience in there. But in terms of domain expertise I'm 10/10. I did statewide policy implementation. I was the expert in my narrow field.

So, when my legislature wanted to fix a problem with pathways to college, some of my work was used in writing that policy, but in the end the policy was a negotiation between the elected officials and it contained some compromises as well as some, well, to be charitable, bullshit that sounded good to voters but either didn't help with the task, or possibly undercut the task. They wrote that legislation and handed down to people like me to implement it. So I got my colleagues together, from all across the state and we got to work on it. We understood the intent of the law, and we agreed with it, but the details left much to be desired. There were gaps - things we needed in the law to really be successful, like mandates on institutions to do certain things, which lacking a mandate we needed to get them to volunteer to do this thing. It was hard to get nearly 200 universities in 3 systems to voluntarily do the thing needed. It took a lot of time and energy. There were also things that we could exploit, sloppily written language that allowed us to violate some of the bullshit in spirit, but comply in the strictest legal sense.

And in this process we were trying to automate as much of it as we could, and we did quite a bit. Some of that came by taking other legislation and exploiting it for our purpose. Some of the previous efforts by the legislation to solve this problem were still on the books, and we leaned into that to compel some of these things to get done. When we needed a classification system which this new law didn't provide, we could have formed hundreds of committees to invent a new one constantly trying to pull them into a cohesive system, or we could take the system the legislature previously tried to implement but failed and simply adopt it, which is what we did bypassing all that work.

But the thing we couldn't do is go back to the legislature and say 'yo, this job would be 10x easier if we made this minor change to the law that we suspect nobody would take issue to'. I mean, we tried, but in a political system which is perfectly engineered to create conflict rather than solve problems, that's a fraught thing to do. Democrats we talked to were afraid that Republicans would use that as a vehicle to undermine other efforts. Democrats we talked to were afraid that other Democrats would use that to rebalance the system in a way that favored their district, etc. So instead we had to do the 10x as much work.

Now, at work, because I was also a policy guy, if we were implementing a system that could be radically simplified by a small change, I could bring that to my colleagues have that discussion (because I was also in the weeds on it, I was very good at explaining it in ways the policy people could understand - that was just practice and knowing how they think and what their talents were) and so getting a minor policy change to make implementation vastly easier wasn't hard to do. Inside a company, that's usually how you do things 'boss, this new product is almost impossible to code up because of this provision, can we get that changed in the following way' and a week later boss has run it up the flagpole and now that's been changed. You can't do that in government - at least not in the 2 party pay-for-every-inch-of-land-in-blood political space that we have.

And note, Social Security is 90 years old. In that political space, they do not clean up their old work very often, so what social security does, and who it serves, and all the little rules and exclusions and interactions with other parts of government have been changed literally hundreds of times and old rules not necessarily removed or brought into alignment with newer intentions. Code has to reflect all of that because the SSA folks don't get to massage it to make it work - they have to implement exactly what was written, and anything they can't gets handed off to a human worker to implement. And in a system with 300 million people, that gets out of hand fast, so a lot of things you might have said 'this happens so infrequently we won't bother to spend 6 months to code up', well, that happens in SSA as well, but a 0.01% chance of happening is still 30,000 people - equivalent to a small city worths of phone calls, visits, etc. It's a lot of manpower, so there's a greater incentive to code that up.

So, until fairly recently public workers could opt out of Social Security in favor of state run pension, but only if they took that opt-out prior to a certain date (when the law changed). That meant that the system needed to be able to recognize that a worker was a public sector worker who took that exemption and allow them to not pay the social security portion of payroll tax and only pay the Medicare, etc. portion of payroll tax (this is information that must be exposed to the payroll system of that workers employer, and also must be exposed to the IRS who do the enforcement, so don't break the IRS in the process). If that worker then took a different job outside of the public sector, they would be required to pay full payroll taxes again (which the system would need to keep track of this individual with some years exempted ands some years not exempted), and if that person then went back to work in the public sector job, are they required to pay in to SS, or are they still exempt from the previous exemption? Many of these people are still working, still have that exemption. That impacts what they are obligated to pay in payroll taxes, and what they receive in benefits. And that needs to be maintained, well, at least until all of those people die and their beneficiaries die. That could be another 50 years? Maintaining that code and those rules that changed 30 odd years ago for another half century.

Government systems and agencies aren't like the private sector. Private sector problems are fucking easy mode, even in areas like banking and finance. That stuff is complicated in other kinds of ways, but you always have the option to not do a hard thing (like my mom's bank just not handling trusts - fuck you, we don't want to deal with that).

So if you want to simplify systems like SSA, you do that by changing the political system so that lawmakers are willing to work together to solve these problems, and not look at them as opportunities to break government, to score ideological points, and so on. Good luck with that.

Those legacy effects can linger for a lot longer than that. The last surviving Civil War widow receiving a government pension was in the news just a few years ago (teenager married a 90+ yo Confederate veteran during the Great Depression)

Changes that don't cancel legacy benefits can linger for much longer than anyone can imagine.

 

[Fritzr]

If you assume spherical cows (bug free software with frozen requirements) then yes.

But in reality the SSA systems hasn't been running without change for decades. It keeps getting changed to deal with changing requirements. Particularly those that result of changes to the law!
And the SSA administration has long long deemed it necessary to move towards more modern architectures.
Which they have been doing. Incrementally.

One more thing:
The SSA core software is built on top of a niche, proprietary ecosystem of both software and hardware.
IBM has been around a long long time but... if somehow IBM were to stop making new IBM Z hardware the SSA would be truly fucked.

SSA's situation isn't like the early 2000s when companies had to move from running Oracle on whatever-Unix to running Oracle on Linux.

It's okay, Fujitsu still has a profitable sideline selling Fujitsu brand IBM mainframe clones.

Oops. They've scheduled the end of life for Fujitsu System Z ... End of sales: 2030, End of support 2035.

Luckily, there are also System Z emulators that run on machines as weak as the IBM 5150 (original IBM PC)

https://www.longpelaexpertise.com.au/ezine/OtherMainframes.php

There are other IBM Z mainframe clones that haven't announced planned end of life, including IBM, but it is looking like now is the time to plan migration to a stable UNIX or other current platform that is likely to have a long lifetime as well as migrating to a server based emulation of a mainframe. (Fujitsu is ending UNIX SPARC sales 2029, & support 2034, IBM supports a UNIX OS for IBM mainframes, LinuxONE, since 2015)

Tellingly, IBM now markets both the Z series and LinuxONE series mainframes as "mainframe servers".

 

[raschumacher]

I would strongly suggest, just like with your student loan records, that you print out your earnings statement monthly and make sure everything looks okay. Then keep a copy permanently at this point.

I would love to, if only I could login at SSA.gov. They haven't figured out yet why I can't. Oh, and, if you call them they say that they no longer snail mail statements. Lovely.
Trump, Musk, their worshipers, and all government-defunding Republicans should eat shit and die.

 

[Fritzr]

So you are pro assassinate the president? You think that right now is the solution? You think right now violence is the inevitable solution, and at the same time look at Jan 6th and think they were wrong?

Sorry, but what you are saying here is we can't win politically so we should start killing people. That's abhorrent. I am not saying we shouldn't fight. I am not saying if they attack we shouldn't fight back. I am saying that the every time someone has taken a shot at Trump it helped him, and that your jump here is extremely close to the people you are saying you want to fight against.

Becoming evil to stop evil isn't worth it. We have a ton of options right now, and should use them and wandering around espousing violence just give those in power a bigger reason to shut everything down and use the power of government to be violent.

Don't be violent. Fight.

The reference is to a classic quote
“war is a continuation of politics by other means”
--"On War" Carl Von Clausewitz (1943).
Recommended reading for military officers and senior politicians everywhere.

It is often interpreted as including minimal targeted violence such as eliminating only the problem without causing injury to others (such as their military who are responsible for preventing enemies from being a threat to the leaders)

Fortunately for Mr. Trump and other national level politicians around the world, this is considered an international felony. The correct way to do it is to send in the military to kill off their military, then exile the problem leader to a life of luxury in a 3rd country It can get messy when it is the leader's own citizens doing this.

In practice it is normally done by deniable assets, as exemplified by the sudden rise in suicidal tendencies among opponents of President Putin.

 

[Fritzr]

I would love to, if only I could login at SSA.gov. They haven't figured out yet why I can't. Oh, and, if you call them they say that they no longer snail mail statements. Lovely.
Trump, Musk, their worshipers, and all government-defunding Republicans should eat shit and die.

If you haven't yet done it, you need to create a Login.gov ID and use that to log in. SSA has shutdown the other log in options.

 

[maxplanar]

Prescient article, it seems.

https://finance.yahoo.com/news/social-security-website-crashes-agency-212827914.html

 

[Fritzr]

Clearly they are too young and inexperienced to have learned rule #1: it it ain't broke don't fix it.

But, but, but, their language of choice is unfashionable and that CANNOT be tolerated

 

[Fritzr]

Quite possibly IBM big iron. Which tends to be fairly maintainable as long as IBM is still producing the parts.

And still has engineers capable of doing the maintenance.

Even better, IBM is still designing modern versions that use less power, compute faster, and are essentially drop in replacements for 1980s era mainframes.

 

[Fritzr]

I never implied that people who have valid SSNs--regardless of citizenship status--were committing fraud of any sort. How many of the people who overstay their visa end up receiving benefits, though? My assumption would be that almost none do, since it's rather tough to get the SSA to pay out if you can't prove your citizenship status. This is the same thing that makes people using fraudulent SSNs unable to receive benefits. You can't just show up with a social security card and get a check--the forms you need to fill out require you to prove your citizenship status, among other things.

I should've worded my suggestion differently--by giving them SSNs, I actually meant "hand out citizenship like it's candy". The pathway to citizenship needs to be way simpler and way faster. Even 6 months is absurd.

Also, how the fuck did we come to call people who have expired but otherwise valid documentation "undocumented"? I realize illegal was and is a terrible way of describing it, but we don't call a person driving with an expired license an undocumented driver, so why the specific term "undocumented immigrants"? Why not simply "people with expired documentation"? Sure, it's more words--but it's tautologically accurate and doesn't dehumanize. Maybe I'm being unreasonably pedantic for this forum, but I firmly believe that people-first language is a powerful interrupt to the (often discriminatory) mental pathways that shortcuts like "undocumented" help to reinforce.

They are undocumented because they do not have documents proving permission to be in the US.

It is a catchall term that includes "out of status" (entered legally, no valid visa when caught) and "uninspected" (entered illegally, skipping the Port of Entry document inspection)

Not all "out of status" are illegal. For example the 90 day fiancé(e) visa grants permission to remain in US as a legal out of status alien if they married their US citizen petitioner within 90 days of entering US. There are other exceptions to "out of status"="illegal" as well, but there are no exceptions to "uninspected"="illegal", until such time as the immigration system grants a visa (asylum mostly, but crime victim/witness and spouse of US citizen who has filed for a waiver and spousal greencard are two of the other possible conversions to legal presence)

Undocumented is useful shorthand for present in US in violation of US immigration law that excludes those that do not have a valid visa, but do have documents showing their legal presence.

Drivers with an expired or suspended license are referred to legally as "drivers without a license". They can be charged with "driving without a license" when caught. Sometimes there are separate charges for "driving with a suspended license". For the purposes of arrest, both cases are "driving without a valid license"

This is reflected in the immigration system where overstays and uninspected are treated differently by the courts. For the purpose of arrest there is no difference; they are undocumented.

 

[Sjoerd Verweij]

1) 2) 3) 4) 5) 6)

Yo, umm, dog maybe next time we split the difference in research? DM?

 

[Sjoerd Verweij]

This is reflected in the immigration system where overstays and uninspected are treated differently by the courts. For the purpose of arrest there is no difference; they are undocumented.

Last time I checked, by far the most common way for people to enter into a permanent "illegal" status in the US is overstaying a tourist visa, quite often whilst coming in on an airplane. The people swimming across the Rio Grande are not some wacky unknown group, they are asylum seekers and the overwhelming majority happily shows up to their day in court.

 

[Fritzr]

This. Any non-US resident who thinks Oceania won’t encompass them is an idiot. A fascist USA will tag team with fascist India, China, and Russia to carve what remains up.

I do have faith our colonial colonial cousins to pull out of this nosedive. What bothers me is their own lack of faith in themselves. That’s the Con at work now.

Oceania is already working to overrun this expat.

Philippines facing China's claim to anything in the South China Sea. China has declared ownership and this is considered a superior claim (by China) to the court order that denied their claim
https://www.uscc.gov/research/south-china-sea-arbitration-ruling-what-happened-and-whats-next

The 9 years since has just seen increased aggression by the Chinese Coast Guard and military reserves. Japan, South Korea, and even Australia have been added to the neighbors being threatened. Mr. Trump's ¿protection? of allies is simply delightful for President Xi who needs something to distract Chinese citizens who might otherwise worry about the current economic failure.

 

[Sjoerd Verweij]

Java is still being developed.

Project Panama: Interconnecting JVM and native code

Compared to C# and the .NET Framework, Java and JVMs are less open (as in free as in beer open), slower, and lack a list of features the size of the Burj Khalifa at this point. Seriously, it is not even funny. If you told me to backport the source of my current projects back to Java-level language features I would smile, nod, and then I would fucking quit.

Last time I checked Intel will still happily sell you licenses for their optimizing Fortran compiler and AVX primitives, too.

Edit: all I want is proper and injectable extension props. If you don't know what I mean by that, you're probably a Java programmer and you'd probably have an aneurysm halfway through someone explaining