If the rewrite is done by people who don't understand the current system any slim chance of success evaporates. and these dumb fucks keep showing us they don't understand much.

given how far behind he is on everything, I'm excited to hear about how he's putting all government servers on the MEAN stack

This type of huge conversion is (as it sounds like you know) literally one of the unsolved problems in software engineering. Many such projects fail to varying degrees, despite large investments, even without [such] a bunch of buffoons in charge.

For an example, the IRS has been in the process of converting its Individual Master File processing from Assembly Language to Java since around 2009. As far as I know, it's still not done but they've made some progress.

Ah that makes sense because there is definitely part of me that thinks that the overarching goal - i.e. having key governement IT systems work in a code that lots of people under the age of 65 can use - is fundamentally a good idea!

Tough part is, you might need to keep doing this every 65 years.

If someone says "convert from COBOL in a few months" we should give skeptical cracks of laughter and go on about our business. Because none of us were born yesterday. The pity of it is, we can't look away this time, because this matters to more than just one company's stakeholders.

There is also the fact that, in this case, the migration from COBOL is unnecessary. I'm sure it's expensive to maintain. I'm sure it has maddening inefficiencies. But has it done a megalapse? It has not. People get their checks. Numbers balance.

my husband is one of the handful of people who is an expert at cobol and his entire job is traveling around fixing it when some dipshit fucks with it. he gets paid very very well to do this

When I was a CECS major over 30 years ago, I was told if you want job security learn COBOL.

Migration from COBOL probably does need to happen* - and it will require a years**-long effort to do it. *Mostly because the number of people who understand COBOL diminishes each year. **Decades?

I wonder how many people will "accidentally" lose their payments?

"Migrate off COBOL" is a phrase that would give any coder nightmares, and the nightmares will be written in buggy COBOL.

Kind of like confidently marching into Russia in June, only to get bogged down, then winter hits.

a library was getting rid of a bunch of books on COBOL so I dressed as a rabbit and gave people COBOL books as the Cobol Bunny lol wacky teens

It’s the computing equivalent of “invading Russia in the winter.”

I have been assuming they want to improve the SSA codebase, but maybe that’s wrong. Like everything else, maybe they just want to tear it down and make it fail.

brb just gonna migrate off of Euclidean geometry

And Arabic numerals.

I honestly had a great laugh when I read that

When I was a lad, I used to wear a t-shirt the read "PL/1 - The UnCobol". You can easily tell who won.

The VA Hospital in DC took a geological age to migrate off of MUMPS. (Actually, I don’t know for sure if they ever did. Haven’t chatted with my dad about it in forever, and he retired.)

It's like migrating off English.

COBOL needs to be retained for the same reason our nuclear missile silos still use cartoonishly large floppy disks. The esoterica of it makes it more secure.

security through obscurity doesn't work - especially for something like this that's going to get specifically targeted

The fact that these adderall popping 20 year old DOGE kids have never heard of, let alone know how to program/hack COBAL isn’t a bad thing

It's never a bad idea to additionally use Security through obscurity. It's always a very bad idea to only use security through obscurity.

We’re going to call the new system “Barbarossa.”

this shit is why rome fell, for realsies.

Rarely is the question asked, “Is COBOL actually the optimal tool for this job?” It was, after all, designed to be the optimal tool for this job.

Pretty sure the IRS has been trying to do this for years, if not decades. So much that can go wrong.

About a decade ago a local freight railway network spent 18 months migrating from "legacy systems" (in-house software written in COBOL) to SAP before announcing the rollback of the project and a $68M write-down on the books for their efforts. Now that's efficiency.

They're lucky it only cost them $68 million and not their company. THERE ARE NO SILVER BULLETS, as I think Fred Brooks wrote many years ago. (If it wasn't him, I'm sure someone out there can correct me!)

I've done this project successfully ...at a smaller scale ...over several years ...with over a decade of experience ...and an experienced team ...and exactly zero AI they're gonna break social security lmao

It is possible and can be "automated" to a great extent. DWP (UK), migrated from COBOL/IDMS (codasyl DB) to Java(?)/Oracle. It was a "success" BUT it was a lift and shift and still had COBOLness and IDMSness embedded such that to maintain it you needed to understand the old system and languages

The measure of success was that the new system did the same as the old, verified by months, yes MONTHS, of testing. So off COBOL, in a few months, is highly unlikely, and a potential disaster

1. “Efficiency” isn’t their goal. 2. Do everything humanly and inhumanly possible to block this til they’ve seamlessly converted and maintained the efficiency of a comparably sized LESS ESSENTIAL system.

These are people who think the whole "Russian cosmonauts tried to write in zero-G with a pen. Americans gave their astronauts a pencil" story is true.

man, that second sentence in the preview is like getting killed twice

I mean, if they break social security it’s over for this administration so maybe…light this candle?

Sounds like just the kind of project you'd want to have forced your longest-serving staffmembers into retirement before embarking on.

I think only Musk is smart enough to quickly learn COBOL and personally make this effort a success, so he should put himself in charge of this and make his face the face of the initiative. Also he should be able to rewrite the system in 3 months or less, otherwise he's an idiot.

It took me a few lines to realize the sheer malice inherent in this suggestion.

It’s like a well balanced vinaigrette

We don't know it yet because of how relatively young COBOL is, but mark my words. COBOL will outlive all of our bloodlines.

The B52 of code.

Oh my god they’re going to use GitHub copilot to do this aren’t they? 🥲🙃

Grok

Like AI was trained on COBOL 😂

😂 So true.

Bro, it’s gonna take YEARS to get off of COBOL. Forget MONTHS. And with generative AI? It’s gonna get stuck in a jump loop and never get out. They just want to throw a grenade in the computer room and close the door, don’t they?

they’re gonna test on production and ruin everything forever

They ultimately don't care if they ruin everything forever. The ruining it is the point. For all of my life the Republican playbook is to tax cut, legislate, or in this case sabotage government programs, then use their struggle as evidence govt cant do anything right and should shrink.

Admiral Hopper should give 'em a whiff of grape.

No coding task "requires" the use of generative AI. I use AI to help me autocomplete tedious code and it still occasionally generates blocks of utter horseshit.

I use it only to give me a better understanding if it's something new to me. Not often as 40+ years as a dev has given me a lot of exp to draw on. I have yet to see a code generator that hasn't had it's end product require countless man hours to re-write the abomination that was created.

I declined to extend my last BI consulting gig because a new manager came in, selling himself as a VBA automation guru, and from the day he started 50% of my workload was cleaning up the dogshit that chatGPT was generating for him.

YIKES!!!

Oh, ! That's a dreadful way to have to earn your living.

Illegal considering access not granted violates multiple laws. Secrecy laws. Sovereigns have data on that system. Illegal for Nazi access.

Right next to land war in Asia, you are right

I wonder if one could make a case that "staying on COBOL" is the long-term cheapest option, all alternatives considered. Even if you have to do things in other systems that ideally should be done by the COBOL system (but it can't). Even if you have to train and develop your own COBOL programmers.

Technical Afghanistan

""generative AI"" welp, there goes everyone's money

It's a lot easier when you don't care whether or not the system you're tinkering with continues to work correctly.