Easy to be irresponsible with other peoples money.
We want government to be cheap. You get what you pay for.
Sure, but it’s not like the consulting firms are paying their “contractors” more, they just siphon up the difference for their shareholders.
That’s exactly how it works, look at contract government salaries compared to anything in the private sector. They charge the government more as rates “go up”, but that certainly isn’t passed along. If large contracting companies really offered value to the government _and_ kept up with market rates for their employees, the state of federal software wouldn’t be what it is today.
I rejected the idea that the consultancy would get a rate increase based on "market rates being higher" and then just capture it all - in my (admittedly somewhat limited and path dependent) experience that just isn't how it works. It's more like we pay randmeerkat $X, we bill them out at $X * factor + overhead. "Market forces" mean we have to go Y > X in renewal or we will lose randmeerkat & friends, so now they get $Y and we bill $Y * factor + overhead. It's of course usually more complicated in general, and overhead especially likely isn't that simple.
Nowhere in there is the assertion that X or Y is what randmeerkat would get on the open market. Importantly, their market isn't really "programmers", but "programmers that work in govt + contracting halo". Which is part of why the idea: I could get $N more in SF tech may be compelling for you, but isn't compelling for them (unless too many people actually make that change, instead of just talk about it).
Also there are many other ways they can raise there rates, but if the claim is that it is due to market on the developer salaries that are going to be a line item, there is going to be at least first pass look at a) is that true (find some market data and wave hands) b) did it actually get spent that way (may come up in an audit).
So the real answer seems to be not "They get more and I get nothing" but "I get a bit more, and they get more scaled by what I get", i.e. "win-win".
There is lots wrong with the system of contracting, but I don't think criticizing a cartoony straw-man of it gets anywhere useful.
What is definitely going to happen is you will attract the cowboys who have convinced themselves and can convince you to pay top dollar. They will come up with lets do kubernetes + service-mesh + multi-cloud hybrid + blockchain + crypto + AI/MI on this COBOL ... They will take top dollar, make a bigger pile of mess, sell it as a success story and move on make bigger messes.
Highlighted in the sibling comment: "you will get whatever candidates decided to live in Mississippi"
As if top talent never decides to live in Mississippi and accept 250k/year as their salary ceiling...
It's not impossible and $250k is nowhere near the ceiling. But you're not going to find those jobs on a causal LinkedIn search like you would in California.
It can attract people who are already good software developers absolutely, but the second point being made is that you also attract a lot of charlatans as well and so just throwing more money at the problem isn't exactly a great solution.
You need a solid culture to go with the money.
I find it hard to get roles in the midwest (large metro in Ohio) clearing 200k, though sometimes bonuses get you up there.
I always assumed the reported 250k plus salaries were fake news or only reserved for children of powerful people.
$250k total comp isn't even particularly high at one of the big companies.
It definitely isn't easy to find those numbers if your CoL doesn't require them. You'd need to move to those areas.
I'm not a sack of chemicals whose desires can be reduced down to some pithy pseudo-scientific Maslowian quip. The only thing I crave from my job is cash so I can take care of me and my family. 250k/yr for a COBOL position would suit that end just fine. You can keep your autonomy, mastery, purpose, and all the other foofoo, just hand me the check so I can be on my way.
I thought you were against reductionism?
sounds like purpose, of which you need mastery of some skillset to earn your autonomy to furfill.
That said, I'm sure there's much easier ways to earn 250k than untangling a governement codebase. Not that they are offering that to begin with.
>just hand me the check so I can be on my way.
Yeah, you probably wouldn't pass clearance with that attitude if we're being frank.
Fed govt migrations are ALWAYS shitshows. The army alone has like 20 individual email services. The pentagon is to org complexity what big tech is to technical complexity. It’s turtles all the way down.
This is literally one of the largest bureaucratic challenges on earth. There’s no simple fixes
There's a time and place for democracy - but large IT projects are not. Do a thorough need analysis, compare with what's reasonably possible using COTS software, and adapt or discard what is not possible.
This is the funniest possible response. The 'thorough need analysis' is the reason we are where we are now!
The thoroughness is so thorough that the needs change faster than the analysis can be completed, obviating the analysis in the first place.
Ironically, my friend has programmed COBOL on DoD mainframes, and still does as a consultant, having interacted with them for decades. Back in the 1970s she had a surplus IBM minicomputer that her father brought home for business use, and learned COBOL on that, and used the computer for furthering her legal research.
Do you think they aren't? I don't know any now, but the last one I did was making more than that 15 years ago.
Old crufty legacy systems in a big bureaucracy are many peoples idea of hell. Empirically speaking, good pay and job security aren't enough to have people knocking down your door for that kind of work if they have other decent options.
Worked at a big bank migrating some old code to new stack one time:
- Original developers are gone.
- Existing users don't understand the systems outside of their own narrow use-cases.
- Code which was committed to repositories didn't match the actual running code, because people would go into production servers, manually edit the code there, in place, not document or commit the changes to repo. There were like 20 years of adhoc changes made in this way.
We had some business analysts which were supposed to talk to the users and gather requirements - they very quickly simply gave up and asked the developers to 'reverse engineer' everything.Anyway, it was a hilarious shit-show, but kinda fun at the same time (if you're able to ignore incompetent leadership pressure - a skill I mastered during this time).
I remember the first time I saw this at a small company. The head programmer was the only one with the actual production source, despite a working source control system. He complained about how people didn't appreciate how hard it is to make builds work. I should have left when we lost three man months of code to losing email. I only left after screaming at the head programmer for refusing to stop making the repository unbuildable at all.
Not that all software from then was wonderful. Or that COBOL magically guarantees great code.
But, on the flip side, you can't say the code base isn't battle-hardened (to various literal degrees).
* edit: corrected sp
This is why big companies use Oracle or SAP for their ERP, auditors are very familiar with it and the way it functions is well-known and understood.
It was a naval research project headed by Admiral Grace Hopper to program with words rather than formulas like earlier "high-level" "autocoders", enabling secretaries to transfer their business logic into computers. It actually intentionally has a lot of the same structure as a recipe, hoping that this would allow easier uptake by secretaries.
It will just cost twice as much as the team that iterates quickly. It will cost more because I will spend more money on business analysis up front. I will spend more money writing automated tests. Most importantly, I will maintain my own toolchain and libraries or pay someone to do so. Every piece of software deployed will have someone accountable to my business to maintain it.
It turns out that this is expensive. It's no more expensive than maintaining the COBOL code, mind you, and would likely give you better results than the equivalent COBOL code. Most companies will take their chances with less expensive software because that's what their competitors do.
About a decade ago, I saw one of the major systems you talk about up close. While they spent that amount of money for mission-critical systems, they didn't do nearly as good of a job on the average software in the system. I'm not convinced that any business is ready to go back to the battle hardened days.
In my experience, the quick teams only learn the surface content. A lack of depth tends to result in more errors down the road and poor architecture. But I guess it doesn't matter since half your team will jump to the next project company in 2-3 years anyways and take that knowledge with them.
Today, the payroll supervisor has no idea how many processes work.
It will be guaranteed to run in 100 years time.
Any other language can die or have incompatibility issues. Over that timespan.
I am not convinced Node will be here in 100 years.
* assuming that the 100 years hyperbole is accepted
This is never guaranteed. How do you know that nothing will change in JS land in 10 years, much less 100? You don't, that's how.
There's no guarantee we'll even be running programs written in C in 2123, let alone ones written in JavaScript. 100 years from now is a long time.
I'm not sure I would trust these people over those of the past, regardless of what they say they will accomplish, Chesterton's Fence, and the Lindy Effect, and all.
>spend more money on business analysis up front.
Entails. But sure, if the government doesn't want to pay for that quality, they can spend just as much or more resurrecting Cobol devs in cryo-stasis in 2099
Your statement is undermined by the fact that the system has failed its sixth consecutive audit. If it's not hardened against failing an audit, I'm not really sure what it could possibly be hardened against.
Managers need to be careful of Goodhart's law. (in this case "passing inspection" may not be the best target)
The DoD does not.
This is the same code that uses two-digit years and doesn't support lowercase letters?
That serves as evidence against this point, though:
> COBOL in 60s was written with a lot more focus on lasting in general.
I worked on enterprise cobol for nearly 20 years and can generally attest that cobol code from the 60s is absolutely nothing at all what you’re describing.
It’s brittle and constantly breaks, requiring 24/7 on call support for people to deal with issues so it doesn’t hold up the next days work.
It was not written with any semblance of engineering prowess. Most cobol written in the 60s was written by business managers, not programmers.
The reason it stays running isn’t because it’s battle hardened. It’s because nobody in their right mind would ever pay to untangle the unfettered mess.
Now the systems I work on are hot garbage because the tech is constantly needing upgraded versions, changes to the code are onerous because of the way the system was designed, we are constantly having data issues for various reasons, and people jump teams every 2 years so there's no continuity.
I'm convinced it's a leadership issue. Push for speed and you'll get it. It will cost you in other areas.
Something that's interesting to think about is that because the cost of changing COBOL is so high now, only the most important change requests actually get implemented. Compare that to a team that can develop and ship a change the same day; their efficiency leads to more requests that end up being bad ideas. The Pentagon has designed a system to ensure that only the most well-thought-out ideas make it into production, they just went about it in a very strange way.
It can take months for a new developer to understand a legacy code base but an LLM with a big enough context window would be instantly productive.
one advantage the government would have is training/fine-tuning on a hundred million lines of domain specific cobol.
It's also wildly bad to plan to train and fine tune on code that you know has bugs. Already we have Copilot generating code with trivial vulnerabilities because that's what it's trained on.
Honestly I think editing code will be easier than creating wholly new applications to precise spec.
Once you have the detailed specifications the coding is relatively easy. Some of that can be partially automated using AI CASE tools, but that only gives a marginal improvement to overall project productivity.
One of my first one afternoon (for prompt refinement and trying it out) toy GPT-3.5 “applications”—really just a prompt—was having it interview for reauirements and draft and progressively refine a requirements doc in a specified (by reference to an author in popular Agile literature, not by specific templates, so just relying on the model’s general training) format. Its pretty much what convinced me of the broad utility of the technology.
I absolutely think that with a bit of effort, GPL-4-128K, a custom RAG framework and/or the Assistants API, you could build something that handles a lot of this. I’m not really bullish near term on LLMs as full replacements, but I can see it as a big force multiplier.
So lots of tedious repetitive manipulation of language to extract facts and transform them into a systematic format? My prediction: LLMs are going to eat that for breakfast.
Never used it (and hopefully never will), but yes.
In the Navy I worked on the "Zeus" project, it was the most expensive illegal operation in the world. We would drop $1 millon dollar hydrophones like they were candy, $1 billion "tails" we were a little more careful with but would drop and re-deploy those at least once a month.
The whole idea of audits for DOD is nonsense and at best is a distraction. I was a nobody and saw a $trillion wasted on what amounted to a "plausible deniability" project so they could appear to Congress to be doing their best in case NK or Russia ever fired a nuke.
The whole thing is a joke, from beginning to end and there was zero accountability.
It sounds like nothing has changed. Relevant to this crowd, maybe: "Most of the Cobol code the Pentagon uses for payroll and accounting was written in the 1960s, according to 2006 congressional testimony by Zack Gaddy, director of DFAS from May 2004 to September 2008."
"Wallace, the Army assistant deputy chief of staff, says the system has "seven million lines of Cobol code that hasn't been updated" in more than a dozen years, and significant parts of the code have been "corrupted." The older it gets, the harder it is to maintain. As DFAS itself said: "As time passes, the pool of Cobol expertise dwindles.""