Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
United States

'The IRS Says There's Always Next Year' (msn.com) 131

The tax agency again delays a vital software upgrade, at the cost of billions. WSJ's Editorial Board: Taxpayers endure drudgery to file on time each year, but the tax collectors seem less concerned with deadlines. A new Internal Revenue Service database, more than a decade in the making, will be delayed another year. And its cost is billions of dollars and climbing. The IRS told the press this week that it won't replace its Individual Master File until the 2026 tax year, at the earliest. That falls short of Commissioner Danny Werfel's goal of launching a new system in time for 2025 taxes, and the delay could mean another year of grief for countless taxpayers. The file is the digital silo in which more than 154 million tax files are held, and keeping it up-to-date helps to enable speedy, accurate refunds.

The code that powers the database was written in the 1960s by IBM engineers at the same time their colleagues worked on the Apollo program. The system runs on a nearly extinct computer language known as Cobol, and though it retains its basic functionality, maintaining it requires bespoke service. By 2018 the IRS had only 17 remaining developers considered to be experts on the system. The agency has sought and failed to overhaul or replace the database since the 1980s. It spent $4 billion over 14 years to devise upgrades, but it canceled that effort in 2000 "without receiving expected benefits," according to the Government Accountability Office.

The costs continue to mount. IRS spending on operating and maintaining its IT systems has risen 35% in the past four years, to $2.7 billion last year from $2 billion in 2019. These costs will "likely continue to increase until a majority of legacy systems are decommissioned," according to a report last month by the agency's inspector general. Each year major upgrades are pushed back adds a larger sum to the final tab. The IRS usually pleads poverty as an excuse for failing to stay up-to-date. Yet Congress gave the agency billions of extra dollars through the Inflation Reduction Act to fund a speedy database overhaul. Since 2022 it has spent $1.3 billion beyond its ordinary budget to modernize its business systems. Taxpayers will have to wait at least another year to see if that investment has paid off.

This discussion has been archived. No new comments can be posted.

'The IRS Says There's Always Next Year'

Comments Filter:
  • She still works in cobol.

    Then it's gonna be 18 people who have a vested interest in not outsourcing themselves to the cloud.

    • Re:I'll call Mom (Score:4, Insightful)

      by unrtst ( 777550 ) on Friday September 13, 2024 @10:33AM (#64785347)

      Only 17 developers in 2018... been working on overhauling it since the 1980's... $4 billion over 14 years, ending in 2000... $2.7 billion last year, up from $2 billion in 2019... Since 2022 it has spent $1.3 billion beyond its ordinary budget

      If those 17 people aren't multi-billionaires by now, where's all that money going?!??!!

      FFS, hire a LOT more developers! With that much money and time on the line, setup multiple developer groups and have them each go at getting to a solution first. You don't even need to find competent COBOL programmers (you don't want this written in COBOL anyway).

      And yeah... I know that's over simplifying, and those 17 devs don't necessarily account for the current development team, but come on! Billions each year? You could pay $1 million a year per developer and still have over 2,000 of them with $700+ million to spare (for management?).

      These costs will "likely continue to increase until a majority of legacy systems are decommissioned," according to a report last month by the agency's inspector general.

      And they're claiming the legacy parts are what's causing the increases? How has that not been shored up by now? That was a $700 million increase this year alone!
      Also, "... likely to continue ..."!??! Is this a threat of some sort, or just utter incompetence?

      • I see you are new to the government work environment.
      • by korgitser ( 1809018 ) on Friday September 13, 2024 @12:35PM (#64785687)

        The cost is not the developers, Cobol or not. The cost start with the system predating databases and it being in fact a single text file that stores /everything ever/ since the sixties. This setup does not parallelize, so they basically buy a new IBM mainframe every few years to run it, and basically pay whatever the asking price is, hardware, licensing, the whole lot of it.

        There really is no alternative but a complete rewrite, but even though you can start one, there is basically no way on earth one of sufficient quality could be produced by the modern software industry. The rewrite is just a run-of-the-mill milking of the govt cash cow, by whoever bribed themselves to get the job - good for as long as it lasts, then scrapped, rinsed, and repeated from the beginning by another opportunist.

        • There really is no alternative but a complete repeal of individual federal income tax.
        • by clovis ( 4684 )

          Rewriting the COBOL may well be the least of the problem.
          One is that this runs in the IBM mainframe OS utilities environment. Knowing the IBM job control and OS specific calls is another thing.

          There's also that Ii's likely there's a good bit of assembler written for the 360 still lurking about.

          Back then people often designed file record size to fit or be arithmetic multiples of the disk drive tracks in use at that time, not to mention the whole 80 column card thing. Or did they ever switch to 96 column card

        • The "software industry" both is and isn't the problem. If you contract Oracle or whoever to do it then yes, they will fuck you. If you build a team and have them develop the new product, you have a chance.

          The problem is therefore fascism. Corporations have such a stranglehold on us that we can't do it the right way. They buy/make the rules/laws that keep us contracting it all for billions instead.

        • Re: I'll call Mom (Score:5, Informative)

          by kenh ( 9056 ) on Friday September 13, 2024 @09:24PM (#64786785) Homepage Journal

          The cost start with the system predating databases and it being in fact a single text file that stores /everything ever/ since the sixties. This setup does not parallelize, so they basically buy a new IBM mainframe every few years to run it, and basically pay whatever the asking price is, hardware, licensing, the whole lot of it.

          Tell me you are ignorant about mainframes without saying you are ignorant about mainframes.

          There have been mainframe databases for years before the relational model came along. IMS/DB came out in 1966

          The idea that "a single text file contains everything" is just asinine.

          The idea that you just keep buying ever bigger mainframes year after year is nonsensical - you buy additional processor complexes and join them to what you already have.

          And the job doesn't parallel process? FFS, you really imagine it serially process 175 million tax returns one at a time? Mainframes have been running "VMs" since the 1970s, first under the VM operating system, then MVS in the 1980s, and more recently System Z.

      • Hold on now, you're forgetting one thing.

        After doing a Star Trek short they had to follow through with a Next Generation, Deep Space Nine, and Voyager homage as well.

        You gotta have the whole set, right?

      • Also, "... likely to continue ..."!??! Is this a threat of some sort, or just utter incompetence?

        Ummm, this is a textbook case of corruption. Enjoy the show while nobody prosecutes. :)

  • by VampireByte ( 447578 ) on Friday September 13, 2024 @09:58AM (#64785257) Homepage

    If COBOL is nearly extinct then why is over a billion new lines of COBOL written per year?

    • by gweihir ( 88907 )

      Indeed. COBOL may not be the new cool thing, but it is in regular use. Same for FORTRAN.

    • There's nothing wrong with COBOL. But the technical debt they've acquired over the decades is definitely worth a rewrite. It still takes someone with COBOL knowledge to figure out what all the hacks and workarounds are actually doing and to get functionality moved to what's new. If writing from the ground up today, COBOL is not a likely choice just because of the number of people proficient in it.

      Replacing monolithic systems all at once is really complex. If they don't modularize at least some of it, th

      • Most code written in any language before the 80s is unreadable by today's standards because the people writing it came up when people were counting bytes of memory and coders usually worked in assembly, so they took their spaghetti and inscrutable acronyms and abbreviations with them to fortran and c and cobol and whatever.

        Java style guides in the 90s were a direct overreaction against this history.

        So no, there's nothing wrong with cobol or fortran, but it's still a pain in the ass to deal with cobol or for

      • > There's nothing wrong with COBOL

        It is verbose as fuck. Too much signal effectively becomes noise.

        Over time popular words tend to become shorter to minimize the time and brain power needed to communicate and parse them. COBOL is an archaic language designed under this delusion that verbosity is "better". It isn't. Pascal does the same stupid shit with BEGIN and END. C has a good balance between terseness and verbosity which is why it has lasted.

        • It's no surprise that COBOL has been supplanted with so much Java.

      • Replacing monolithic systems all at once is really complex. If they don't modularize at least some of it, then they're going to run into the same problem next time.

        Is the air traffic control system still running on UNIVAC computers?

        Is the FBI still running Windows XP on desktops?

        The federal government can't replace large systems , it just doesn't happen.

        • That's why they have to be eventually replaced with interconnected small systems. If they can ever replace anything.

    • If COBOL is nearly extinct then why is over a billion new lines of COBOL written per year?

      If COBOL is nearly extinct and so many systems use it, then why hasn't anyone written a COBOL to Java converter? Sure it won't be perfect (or even very good), but at least Java programmers are readily available and you can use modern tools on your code.

      • Well, I think it was IBM (later sold off) that came out with EGL and VG 25 years or so ago. It is its own language, and you generate either really nasty looking but working java code (and JSPs) or I would assume really nasty COBOL code and whatever it uses for front end stuff. Ties into Websphere, etc.

        The place I work for is still dependent on it .... it isn't so bad, some of the things make sense (a record data type that can represent rows/cols on one or more db tables), some stuff would probably make mo

      • They did, then they realized JAVA sucks for typical mainframe applications.

        There are COBOL to the latest 'Flavor of the Month' converters, but they don't solve the problem. The problem is the code is the documentation, and for lots of code not only has the original developer died, but their children have all retired, and their grand children aren't interested in learning COBOL.

  • Not exactly waste (Score:5, Insightful)

    by dpille ( 547949 ) on Friday September 13, 2024 @10:14AM (#64785301)
    If I were cutting my lawn with a push mower, making and deferring my plans to buy a riding mower might indeed be costing me something. But it's a hell of a lot better than using scissors the whole time.
  • by JBMcB ( 73720 ) on Friday September 13, 2024 @10:20AM (#64785319)

    Ah yes, the act designed to reduce inflation by dumping another trillion dollars into the economy.

  • by mspohr ( 589790 ) on Friday September 13, 2024 @11:13AM (#64785453)

    The Repubs have been starving the IRS of staffing and IT upgrades for years.
    Finally the IRAct provided a boost in IRS Funding for more staff for audits and IT upgrades.
    Expect these expenses to increase in the coming years. It's a good investment. First six months of increased audits brought in over $1 billion (from rich people).
    IRA funding also allows the IRS to set up on-line tax filing so people don't have to pay for third party tax software. Last year was tested in several states. Will roll out to all states next year.

    • by whitroth ( 9367 )

      Yep. And they've been starving the IRS because *they* want to cheat on their taxes. I saw a report last year, I think, where the IRS was saying it couldn't go after the wealthy, because it didn't have the personnel. to do so. Wonder what that means? Look at TFG's lawyers pushing absurd motions, anything, to delay an actual trial.

    • Re: (Score:3, Insightful)

      by Zak3056 ( 69287 )

      The Repubs have been starving the IRS of staffing and IT upgrades for years.

      If a starvation diet is in the form of 40,000 man years (at $100k per man year) worth of "upgrade" that have thus far failed to produce anything useful, I would hate to see what "fully funded" would look like.

      • I would hate to see what "fully funded" would look like.

        It was fully funded in the 1980s... it was so well funded, that Congress became afraid of the IRS in the 90s. They have since defunded it because if it is strong enough, it can hurt them.

        On the bright side, low level citizens such as yourself no longer have to deal with out of control auditors taking everything you own so you are incapable of even defending yourself. No money, no lawyer, automatic win for the IRS. It was a bleak and grim time back then. Please do not bring it back even if watching Congress

    • The IRS is wasting BILLIONS but some how they are starved for IT funds? How do you square that logic?

      That's right up their with crying about education spending without realizing we're the 5th highest spender on education in the world. I got that info from this link https://nces.ed.gov/programs/c... [ed.gov]

      We don't have a funding problem. We have a spending and accountability problem. We've no accountability so we just keep spending more money thinking that will some how fix the problem.

      Also, it's not hard to imagin

  • Remember when IRS funding was cut by 25% [politico.com] last year? Do you honestly believe that didn't impact large projects like this?

    If you want to blame someone for the delay then you need only look at the congress critters that demanded cutting funding to the IRS or they would cause the US to default.

  • They were learning new tech to replace all the COBOL systems. Just by listening to them, I could tell the project would fail.
  • The tax agency .. vital software upgrade

    Vital tax software, surly a contradiction in terms /s
  • Can they not wrap the old system with another, pass on anything mundane to the old system and handle anything new in the outer new system. Sure they may need to continue to fix some bugs in the old system but as time goes on the outer new system will take more and more of the load.

    The outer system can even have a go at doing some of the old-inner system's functions and compare its output.

  • by Murdoch5 ( 1563847 ) on Friday September 13, 2024 @01:39PM (#64785819) Homepage
    Ignoring the fact it's written in COBOL, replacing it can't really be that difficult, what key element pose a problem? There was a post a few weeks back about a $400 million project being effectively useless, and again, here we have waste on a scale that makes no sense.

    Lets assuming you need 100 developers. You need another 15 IT people, and 10 DevOps, DevSecOps, and aux employees. We'll add in a margin of 25 additional employees for what ever, and all of them make 200k/ year. That's 150 x 200 000 or $30 000 000 / year. What else do you need to add? Cloud? Okay, let's factor in 50k/month for infrastructure, of another 600k/year, With a fairly unreadable, but probably defendable, $30 600 000 / year, being required, even if took 3 years to build the system, that's $91 600 000, so where is the rest of the money?

    If you're given $100 million dollars, and can't deliver, it's either outright fraud or fraud through incompetence. I would love to hear an excuse that will hold up, and make sense for how the transition hasn't been accomplished.
    • One problem might be that these are legacy applications and may not be well documented. COBOL was supposed to be easy to read, like English, so the COBOL code I've seen had zero documentation, you were just supposed to read the code. And that code has been heavily modified over the years. So I would guess there is no documentation of all the things the code does, so the first step would be reverse engineering it to create a spec that could be used to rewrite the entire thing from scratch in another langu
      • Okay, so take a full offline copy, and for a good use of AI, put it to work. I've seen my share of head spinning code bases, so use AI to unwind the mess. Once you have it unwound, which I can grant might be 6 months of work, you start putting to back together using updated frameworks and languages. Actually, using AI again, you could have it transpose the COBOL into TypeScript, GO or Rust, and then, since memory isn't so sought after, as it was in the 60s, document the living hell out of it.

        I'm not jo
  • I've never even heard of COBOL security reviews or best practices. I wrote a little of it a long time ago, it was a lot like writing code in BASIC and while it does read like English, that tended to obscure the actual logic of the language because like any programming language it is important to be able to express intricate details, which creates weird syntax. Does COBOL have network libraries or API's? (Seems it does, see below.) I wonder if the security of the IRS's computing is mainly due to the obsc
    • by whitroth ( 9367 )

      I wrote a lot of COBOL (and fixed it) back in the day. To hack it, first you have to have the code to work from... and a hell of a lot was all written in-house. Second, since late eighties, you can run COBOL on PCs, but how many crackers will do that?

      • by rapjr ( 732628 )
        But many attack techniques don't require the source code, they just probe for vulnerabilities till they find something. If you can run COBOL on a PC that makes it _easier_ to find an attack before trying an attack in the wild, reducing the risk. Crackers not running COBOL is an argument that COBOL is obscure, so it doesn't get attacked, but attacking the IRS and banks is a high value target, surely someone is going to try, like say Russia and China?

Single tasking: Just Say No.

Working...