'Eventually COBOL will need to interface with new code...' Eventually? Do you suppose these programs had support for EFT, ATMs, bank-by-phone, online banking, mobile apps, etc when they were written 50+ years ago?
Do you suppose there won't be new requirements to implement in the next 5 years?
When you say rewriting is a good strategy, do you have ANY idea what that entails?
It entails requirements gathering not just through stakeholder analysis, but through forensic programming. You need to ask engineers how the current systems work, ask managers and administrators what their technical tools allow them to do, piece together a conceptual view of the current system, and have programmers examine existing code--both large spans of barely-readable source code and decompiled and debugged binary blobs for which source has been lost--to get a list of required features, existing systems, and current integrations. Both the goal and the technical details of the current implementation need documentation.
You then need to structure the actual work. This first requires a strategy for incremental delivery, because you're going to need an agile methodology for this--there is far too much risk to write the new system in a vacuum, "test it," and just drop it all at once. You'll need to design ways to test the first pieces you can deliver without messing with production, and to unbolt the relevant parts of production and bolt in the new system. This repeats as you make further releases, replacing your hodgepodge of multi-era code with a unified code base.
You need back-out plans. You need to identify all risks, rank them, mitigate them, build contingencies, and ensure that anything you screw up can not only be undone, but identified. Messing up data accounting in invisible ways will tend to run for a long time before anyone notices the error. This means parallel deployment as a test environment--possibly for months--to validate your outputs on real-world workloads against the existing system before deploying the rewrite.
That's only some of the bare, early technical pieces. That doesn't include procurements, human resources, buy-in, communications plans for literally everything that impacts anyone, management of legal and regulatory concerns, and the like. This isn't a small project; it's a massive effort to remove decades of technical debt and reduce long-term operational risk. It's like the removal of a cancer that's wound its way dangerously through tissue, but is still not fatal and can be cut out without destroying the body: the longer it grows, the harder it's going to be to remove, and the more likely it is to lead to sudden vital organ failure. Banks and insurers are functionally-immortal and don't simply have to outlive their barely-viable systems, like a human who might make 90 or so and die of heart failure before the prostate cancer even starts making them ill; they can actually keep running forever, and die when the system fails.
Your COBOL program is probably running under CICS. The hardware CICS and your program have been running on have been optimized for your workload. Your data is probably stored on ECKD DASDs. Your data is probably stored in packed decimal format, so it can be operated on with a single, optimized, machine instruction.
I've seen data-crunching mainframes in insurance companies running modern code. Not C#.NET, Java, and Python; they've got JCL and Rexx to chew through massive data sets and run huge reports. They use mainframes because the software on the mainframe and the I/O hardware does in twenty minutes the job a $300,000 bank of servers can't do in literally three weeks. They still use software written this century, and they've left off software written back when mainframes were "The Future". That software doesn't look like anything today's Web developers are familiar with--no PHP, no C#, no MongoDB-
When I was on amphetamine, I started using soylent because I could only get 1200kcal/day. The original dose (20mg MAS ER) was so powerful I couldn't eat at all: food had a terrible taste and texture, and swallowing felt like an invasion. It was hard to get down. I eventually settled on 15mg MAS ER; 10mg gave me anxiety, and 15mg only made me mildly-depressed. I break over at 20mg like an avalanche effect, complete with a mix of mild and moderate-severe overdose symptoms. The pharmacological window for amphetamine is supposed to be huge, like 8:1, so a large dose usually makes you kind of high and causes insomnia (any dose causes insomnia; it takes 26 hours to be capable of sleep after taking amphetamine), while very large doses do slow damage and get you really, really high (consequence: you're going to feel shitty later and, eventually, you'll reach a point where you're never going to feel good again).
By contrast, alcohol doesn't seem to work too well on me. Benzodiapezine-likes (Eszopiclone, a non-benzodiapezine) seem to not do much at normal doses. Enough alcohol will get me drunk, but it won't muck about with my reasoning abilities; the whole experience is terrible, involves a mixed manic episode, and leaves me incapable of sleep as if I'm on a stimulant--even at 0.34. One of the annoying things about the emotional clusterfuck is the emotional clusterfuck, which I'm stuck observing because of course I am; you'd think I'd just go along with the ride, but no, emotions are an external factor and not a driving part of my experience of self, and I'm prone to point out where emotions are eroding my judgment and overrule them (largely because I'm not comfortable with the emotional decision--figure that out). Eszopiclone puts me in a daze, during which I notice I feel really fucking high 24/7, and that my memory is iffy; it doesn't help me sleep and it does nothing until several hours after I've taken the pill.
Drugs suck. Somebody get me some better tools. In your case, you need... something. MSG in your food, or a good dose of Welbutrin (...no thanks) or Phenylpiracetam (I like this stuff; it's relatively-safe). PPR makes my reward system work correctly and allows me to experience feelings of emotional pleasure, but I generally don't use it, and I remove it from my schedule whenever I have actual doctors making drug adjustments so I've got a better baseline during treatment. Even when I do use it, I take a lowered dose to essentially activate wanting without activating the emotional side of liking, because feelings of pleasure still register to my senses as being really, really high and I'm not entirely comfortable with the impact on my judgment (which is stupid: lack of reinforcement behavior has destroyed my judgment, and my primary goal is to avoid any and all discomfort, so I'm not really bothered by failure but don't pursue achievement if it means expending effort; I need good emotional feelings, it's a requirement for proper function, even if it's scary and kind of embarrassing to take drugs to make me feel good).
Ask your psychiatrist to fix that shit. Trust me, being broken like that sucks. The alternative requires just as much discipline, and draws less sympathy--people will criticize you more for getting fat than starving yourself to death--and it's still a lot better to have to push yourself into eating less than you burn and maybe monitoring your intake and caloric consumption instead of just stuffing your face on the couch. When your body isn't telling you food is good, life becomes difficult and miserable.
19x19, 13x13, or 9x9. The number of possible positions is 2^361, but the number of possible moves is a lot bigger. The problem is projecting those moves is long and difficult; you have to be able to extrapolate localized influence and identify how that impacts a global strategy in an abstract sense, or else you can't play. A computer can't track all of the possible moves because losing a big position might be less-relevant than a play elsewhere that establishes power to restrict your gains to 1/3 of the board and solidify the entirety of the other 2/3 for the opponent.
The amount of analysis is huge. You can create algorithms to shorten it, right up until a human does something strategically-novel.
It's the poor who are most-impacted. The general middle-class are less-impacted, and some are frequent fliers; the upper-class may or may not be frequent fliers, and don't spend all their time on airlines. The very-rich have private jets. Everyone at every level tries to optimize for their particular case, but may have a less-expensive option (e.g. 1.5% cash back is worth 1.5%; 2.2% frequent flier miles are worth 2.2%).
As well, the upper class are fewer and gain less by spending on consumer goods. They buy more in luxuries, and besides have limited purchasing: even if Warren Buffet shops at Giant with his AirMiles card, he's only buying Warren Buffet's groceries; there are a ton of people with lower-fee cards shopping at Giant, and so the average is impacted almost not-at-all. Meanwhile, Elon Musk might buy some fantastically-expensive diamonds at a jeweler on his card for some hot girlfriend of his, which might have some sort of impact on the price of diamond necklaces poor people can't afford.
Most of Elon's money isn't going through his personal credit card; it's going through his business ventures. Warren Buffet's money is mostly in brokerage transactions, buying and selling securities, so he's essentially using his money to suck your money out of your 401(k) (yeah, you put your money in funds without being as good a trader as Buffet, this is what happens; you should have had a plan, or just bought the 3% guaranteed interest option instead of dreaming of possible 14% returns on high-risk bonds in the "Aggressive Growth" fund). The money to pay these people still flows up from the bottom; the impact isn't catastrophic.
It becomes a vague system when you realize it's non-homogeneous and still works that way. The practical impact is essentially that the group of people shopping at the same merchants transfers money between themselves based on that distribution, which means the cost of stuff at the Save-a-Lot is probably not terribly impacted in the sense that the poor are bringing poor-people spending habits and poor-people rewards (which they also pay for) and getting little to no advantage/disadvantage (a wash). The stuff at the upper-end jeweler whose cheapest ring costs $80,000 is probably more broadly-distributed across various tiers of rewards--and the jeweler probably only accepts credit because rich folks expect to have credit cards accepted, and will pay the mark-up for merchant fees (you try to buy a motorcycle, the dealer will let you put $2,000 down on a credit card--the rest must be cash)--and it's all rich people one-upping each other on who has the best rewards structure. The common intersection is mostly middle-class and poor, and most of us think we're getting something (mostly, we're evading the penalty for participating by getting a 4.5% discount on gasoline, 2.8% discount on groceries, and 1.5% cash-back on everything else--which we pay for whether we collect or not).
It's almost-correct to say that the rewards programs essentially cost nearly nothing to anyone, except that there are people who don't participate optimally and there's still a spread between people whose optimal behavior brings higher versus lower merchant fees to the merchants involved.
In truth, it's partitioned in odd ways, and participation engages multiple partitions to varying degrees, resulting in optimal behaviors which are not necessarily optimal in monetary return. Basically, you don't have a choice how it impacts you, aside from whether or not you play along to minimize the cost to yourself.
The business owners were prepared for it. Somebody changed the rules. Now they have to have a $250,000 back-up tractor that's going to sit around not doing a damned thing, rusting out, and becoming obsolete. They won't be able to use it once it's obsolete because they can't replace a part that breaks; since that part's not made and John Deere won't repair the tractor anymore, they need to junk it out. Farmers want a 20% profit margin, but the market only gives them an average of 10%, sometimes 8%; that means that one piece of equipment represents profit off $2,500,000 of revenue, and some farms are quite small.
What do you do if a change in business operations at a supplier can have that kind of impact against an operation? What does CareFirst Blue Cross do if the failure of a RAID drive in a server results in the server shutting down for 2 weeks until a Dell technician comes out with a hard drive, replaces it, and resets the RAID controller so it knows it's okay to rebuild the array and turn the server back on? Note that, today, a failed disk in a RAID array doesn't cause a service interruption, and Dell will have replacement hardware in your hands in under four hours; an immediate halt of service and a two-week lag time to get a new drive installed would be a major change to business operations.
We spent $160,000 here over a few months to analyze the results of an upcoming event. We have some really smart people who determined who would buy what and how much they'd pay, and projected the most-likely outcomes so that we could position marketing and take actions which would cause outcomes worth several tens of millions versus just entering blind using a glance at historical data and some expert judgment.
Do you honestly think banks don't do that kind of simulation? It's short-term, multi-variable statistics calculating the probability ranges of a market, not long-term economic models trying to figure out what businesses or sectors will be strong or how the global market will grow across the next 20 years. Banks don't buy this shit in bundles and lose 40%; they purchase what their customers accrue, likely on a quarterly purchase, carrying over what's left from last quarter.
Customers might handle things less-well, and they pay for the privilege. On the other hand, I'm pretty sure that frequent fliers are actually getting frequent flier miles; they might not be using them in the most-efficient manner, but they're getting them and using them.
Sleep restriction. Take your total sleep time, set your bedtime by rolling back from your desired rise time. When your sleep efficiency over a week reaches 90% or more, move bed time back by 15 minutes. If it falls below 80%, move it forward.
I'm an insomniac with a prior ADHD diagnosis. I got Modafinil from a psychiatrist after my attention issues became asinine; it was fucking awesome for 2 weeks, then I got hit with sudden suicide-grade depression. Modafinil doesn't interrupt my sleep; and since I was sleeping 2 hours a night for over a year anyway, the consequences of abusing Modafinil to stay up for weeks on end suddenly became clear to me--which is actually great, because I wasn't able to find that in the literature. I also got a sample of suicidal impulse, but not much; the experience of that kind of extreme depression was too-new and carried too much risk, so I shut those thoughts down as they formed. I'm too hardened now for that to happen again, so I won't get a second chance to examine how the actual desire to kill yourself forms in a safe environment. Oh well. I get why people bitch about depression so much now; I learned to make it go away, but it was never that bad, holy shit.
That lead to a huge run of psychiatric experimentation. Amphetamine for the anhedonia? Makes me depressed (yes, it does the opposite of what makes people want to smoke meth). Melatonin? No dice. Wanted to avoid a GABA drug, so my psychiatrist gave me Belsomra (Suvorexant), which worked amazingly: instead of 2-3 hour sleep latency and lots of being awake for hours at a time in between, I got to sleep in 20 minutes, despite not actually feeling any sleep urge--I needed like 20mg of Melatonin to actually sleep, though, even with the Suvorexant. $300/month so insurance wanted me to try GABA drugs, so I tried Eszopiclone; it made me really fucking high for a couple weeks, then I stopped taking it because I nearly drove into another car 20 hours after the last dose; the withdrawal was hilariously bad, but lasts only a day.
I'm on Atomoxetine now. The drug hit me like a hammer, and 80mg/day was making me manic, driving my heart rate up by like 30 points, and causing ludicrously-high blood pressure. 60mg still hits me with fatigue, so I want to split the dose and cut it back more. Since I started taking Atomoxetine, I can control my insomnia with little more than sleep restriction, although 2mg of 8-hour continuous-release melatonin helps, and an overdose on B6 helps immensely (B-100 supplements have a toxic level of B6; B6 toxicity only shows up in literature after a minimum of 1.9 years of supplementation, mean 2.6 years, and can cause permanent psychiatric damage. Most "dietary supplement" sleep drugs--unregulated in the United States--contain high doses of B6, e.g. any Melatonin pill will contain 5x the RDI, and ZMA contains a lot of B6. ZMA is also essentially a GABA drug like Valium, since Magnesium binds to the Benzodiapezine site on the GABA[a] receptor and induces a similar anxiolytic effect).
So yeah, it's Atomoxetine and sleep restriction, at a minimum, for me. It's fantastic. Meanwhile I see idiots on Dimaxion and Everyman thinking they're performing better because they have less sleep, and I'm like... I remember I didn't realize the sleep deprivation was actually causing me a problem, even though I knew it was miserable. My brain started shutting down major facilities to compensate near the end.
I've considered that Ramelteon might make my life a bit easier, but I don't want to introduce a new dependency into my life right now. I'm curious, because it hasn't been directly-compared to Melatonin in literature yet. (Ramelteon has a 1-2.6 hour half-life; I get no sleep urge. Sometimes, I'd just like to feel like sleeping when it's time to sleep, instead of feeling like I'm dying of sleep deprivation but still not feeling tired and having sleep be just a damnable chore. Imagine if you only had to eat because you'd become dizzy and weak eventually, and food wasn't really that palatable; eating would be terrible, but necessary.)
It's a tractor. You sit on it and ride. It's not a stationary machine or a commercial jetliner.
We're not talking about an extra $10k; we're talking about the tractor sitting around non-working while enough time passes to call that $10k per minute because it leads to a several-percent drop in yields on a million-acre farm. It's not as direct in farming as it is anywhere else: if you get back on in X time, no big deal; if you have to wait 2 weeks, everything goes to shit. An afternoon of annoyance and some overtime is a lot different than losing a dozen days out of a 100-day growing season.
It's a good strategy, and carries different and more-immediate risks. It's an attractive option for me because it allows factoring in new lessons-learned and it makes risks controllable: old COBOL code might not respond well to new requirements, and will eventually need to interface with new code--either new COBOL code or new code in some other language. You can't control those risks as well as you can control an elective rewrite, because the risk of new requirements is often surprising and under time-pressure, while an elective rewrite to remove technical debt is at your best pace unless one of those risks suddenly appears--in which case, you may decide the rewrite isn't ready, and that shoring up the existing mess a bit longer is less-risky.
An elective rewrite requires accepting those risks now and devoting the resources to the project immediately. That translates to "take an active step in causing disruptions and expenses that we don't need to", versus "take an inactive step in causing disruptions and expenses at a future time when we can't avoid it". This form of proposal requires reframing.
Reframing is a known method for improving decision making and reducing dorsolateral prefrontal cortex activation when making decisions. Generally, to make a difficult decision, the dorsolateral prefrontal cortex must activate on prefrontal reasoning and override the mid-brain response (the simple, automatic response) by force of expended energy. For a frame, "Do you want $5 today, or $10 in 6 weeks," you want $5 right now; it takes a lot of effort to resist the immediate reward. A null-framed form, "Do you want $5 today and $0 in 6 weeks, or $0 today and $10 in 6 weeks," produces a decision to wait for the $10 in nearly an additional 40% of the respondent pool. Essentially, the prefrontal cortex sees these proposals differently, and the midbrain has a reaction to both explicit conditions rather than a lack of response to the implied null conditions--and obviously finds getting $0 at any time offensive versus the alternate, so is somewhat less-focused on time as a factor.
So do you want to take controlled, manageable risks now and pay an expense to avoid uncontrolled, time-pressured risks in the future; or do you want to do nothing now and face uncontrolled, time-pressured risks at various points in the future when implementing a long-term solution to avoid more such risks suddenly cropping up afterwards will not be immediately viable?
An event which will occur at a random, probabilistic period and incur some penalty (call it $1,000,000) is less-desirable than an event which will occur at a random, probabilistic period and incur some smaller penalty (call it $100,000). The cost to replace these events--to mitigate future risks--can be high (e.g. $20,000,000); yet those future risks may be more variable ($10,000-$100,000,000, mean $1,000,000) versus the alternative ($15,000-$250,000, mean $100,000). Frequency may also be a factor, wherein a new system may already have the flexibility for many upcoming new requirements, and any changes may bring with them opportunity to cover additional requirements without gold-plating (e.g. it has to handle PNG images, and you used a library that handles 47 types of graphics formats--when it has to handle TIFF embedded in EPS, it already does that, and nobody went out of their way to spend additional time making that happen). Making the system ready for commodity environments means we can build test environments and more-readily detect defects before production changes, as well, without expending millions of dollars supporting said environments.
It's not a simple problem. It requires a reframe to present it as a complex but important consideration, rather than a scary hobby to keep someone busy. After that, you're ready to sit down and analyze the problem in its full scope to decide if you should proceed. I've had times when the answer was no, and that's usually when we're migrating to a new system soon an
The cardholder doesn't pay; everyone pays. The cardholder with the highest-merchant-fee card is not paying that merchant fee; it's averaged across all transactions, including cash (0% fee) transactions. That means cash holders pay the largest mark-up and get nothing; while high-merchant-fee rewards cards with no cardholder fees pay the same mark-up (read: no additional cost over just paying cash) and get the maximum return (read: what's purchased by those merchant fees is paid for mostly by all other consumers).
Everyone pays into the system. The guy with the best rewards comes out ahead; the guy with cash comes out behind; and the guy closest to the average-fee rewards card essentially gets a wash.
Actually it would seem the extra merchant fees are diffused through merchant pricing, as the customer doesn't pay extra to use his card. For example: if fees of 1.5%, 3.5%, and 2.2% show up, averaging 1.8% of all purchases (including cash transactions with no fee), then all goods sold by the merchant get marked-up by 1.8% to include those costs. That 1.8% pays for the frequent flier miles (3.5% fee on those cards).
Those miles are purchased with consumer after-tax money, and taxing them as additional income would be double-taxation of consumer income. QED.
By the by, the 1.5% cash back you get on your card is also paid for by... well, the rest of us who have lesser cards with lower rewards, or those paranoid schizophrenics who only use cash to keep their movements hidden from the gub'mint.
Frequent fliers are, by definition, flying frequently. The summary seems to suggest that frequent flier miles are a huge free hand-out to airlines; yet aren't the frequent fliers actually consuming those miles, rather than letting them default and thus paying for them (or, well, diffusing the price through consumable goods, thus having everyone pitch in a little to pay for them) and not using them?
Upstart and traditional SysVinit pretty much ignored output; if your script didn't send things to syslog or custom logs, then it went nowhere. Pretty much the first thing I noticed about systemd was how much less-frequently I had to modify init scripts to figure out what was wrong.
Mostly my dislike of Mir was because of its NIH bullshit that I got tired of with Redhat already. Wayland shows up, Canonical is like "well we'll write our own!"
Pretty much, Canonical claimed Wayland is inadequate in many ways and Mir does things a whole lot better; a bunch of people (including the Xubuntu developers) evaluated running on Mir, and found it shrugworthy. We've all basically decided Wayland is where everything is going, when everything does eventually get around to getting off X11, and so Canonical is wasting programmer time that could go into improving Wayland instead.
As an example: Wayland has an XWayland client compatibility layer and a weston-rdp compositor. Wayland can seamlessly transition to different compositors, meaning it can draw on your graphics card for a while and then swap over to weston-rdp. XRDP required an entire Xorg server compiled as an RDP backend, and is defunct. It's possible to muck about with XRDP to get Sessman to launch weston-rdp with XWayland; with a bit more effort than that, you can get a display manager running on tcp/3389, and allow selecting (and displaying) log-in sessions over RDP. Basically nobody cares, and instead we have people busy A) just trying to get Wayland up to production-quality; and B) making Mir because Wayland was not invented here. More people could work on problem (A) or (C) if they skipped problem (B).
With Unity going away and Mir probably vanishing with it, that's years of programmer time lost to defunct garbage that never gained any traction.
The software breaks every time you make a change. Need to pull out the transmission and replace a damaged gear fork? The software now requires an authorized technician sign off on the repair, and disables the tractor.
The problem is it cost the farmer $500 to do the repair on the previous model themselves, in an hour. The new model also costs the farmer $500 to do the repair themselves, in an hour; however, the tractor won't "go" until they pay a technician an additional $1,500 to drive out, wave their badge at the device, and whisper the secret word into its ear.
Part of the complaint is they can't fix their tractor and get back to work; they have to take a relatively-significant hit to productivity and put their farm at risk waiting for a service call.
So global statistics show population growth is slowing even though we're still growing population faster than ever because population won't really stop growing until it stops growing?
History has shown, and continues to show, that advances in technology reduce scarcity, and that reduction in scarcity directly causes a population boom. Reduction in scarcity is wealth. It's the capacity to feed 7 billion people on a planet that can sustain 0.63 billion humans. It's the capacity to have cars, roads, and rail lines when medieval society couldn't build all that shit in 100 years even if given the industrial population we have. It's the capacity to afford a wardrobe of clothes--you know, rich people things, since people who aren't the Nobility can't produce enough to make it economically viable for most people to own more than one or two sets of good clothes (which eventually get torn up and used as rags, because throwing things like that away is ludicrous).
We used to have a revolving door, but I'm working with coworkers who have mostly been here for 3-5 years now. They're all cognizant of the market; we regularly talk about how we can't hire additional people because we pay way below market rates--which includes our pay, considering some of us have been offered 80% pay increases at other jobs doing the same shit.
Some of us are stable. Others just don't care to deal with a job change. A few thought the job would be interesting and now are afraid of jumping ship without 2-3 years to show they're not job-hopping. Plenty are placated by the continuous change in our environment and have a perception that things are getting better, never mind that they're not getting paid more. We're all roped in for various psychological reasons that are quite clear to me.
I'm working at a business with low pay--where the average for a programmer is $96k here, programmers make $74k. The same is true of most IT staff, running a good 20%-30% short of the industry median.
We're also fairly diversified and have chicks and people from all over the world in our staff, and have had folks who speak Russian or obscure Indian dialects as a primary language in prominent technical positions. They're also poorly-paid, although near as I can tell we all have about the same salary.
It seems like a form of posturing: we don't want to pay salaries, so we create a perception of... something. We're a good place to work because of something something benefits diversity open-door-policy.
Are these studies by industry, region, experience, and business? Do we say that black women earn 55% as much as white men, or do we say that black women at business X in job Y earn 55% as much as white men in business X at job Y? What happens if business X mostly hires white men for job Y, and business X' hires a higher proportion of black and asian women for job Y but also pays like shit even if you're a white man?
The researchers in lab coats are making big money. A lot of it is their wages. The wages of people at chem synth companies, mining facilities, and other supply chain services is also a big part of the majority cost.
The question is: what fraction of the cost did that overhead reflect? I've seen gross-mismanagement, and I've seen stuff that's so damned expensive that a 0.5% extra expense is... a lot.
Executive salaries are typically interesting in that respect: the actual cash compensation for most executives amounts to pennies per hour. The CEO of Ford gets, from revenue, about 1.05 cents per employee per hour, or $21 per year; stock and options compensation come at the expense of diluting stock--basically the business pays nothing, issues new stock, and investors in the secondary security market all wind up holding pink slips they can sell for slightly-less than they could yesterday. That means Ford basically robs a penny here and there from your 401(k) to pay its CEO, CFO, etc.
People see executives with $20 million compensation and talk about how they should give that to their employees; then you find that their cash compensation (salary and bonuses) is only $7 million (most of which covers taxes on the other $13 million), and their 300,000 employees would each receive $23/year or a raise of 1.17 cents per hour.
So are the R&D perks 40% of the budget or 0.4% of the budget?
Slashdot Mirror
'Eventually COBOL will need to interface with new code...' Eventually? Do you suppose these programs had support for EFT, ATMs, bank-by-phone, online banking, mobile apps, etc when they were written 50+ years ago?
Do you suppose there won't be new requirements to implement in the next 5 years?
When you say rewriting is a good strategy, do you have ANY idea what that entails?
It entails requirements gathering not just through stakeholder analysis, but through forensic programming. You need to ask engineers how the current systems work, ask managers and administrators what their technical tools allow them to do, piece together a conceptual view of the current system, and have programmers examine existing code--both large spans of barely-readable source code and decompiled and debugged binary blobs for which source has been lost--to get a list of required features, existing systems, and current integrations. Both the goal and the technical details of the current implementation need documentation.
You then need to structure the actual work. This first requires a strategy for incremental delivery, because you're going to need an agile methodology for this--there is far too much risk to write the new system in a vacuum, "test it," and just drop it all at once. You'll need to design ways to test the first pieces you can deliver without messing with production, and to unbolt the relevant parts of production and bolt in the new system. This repeats as you make further releases, replacing your hodgepodge of multi-era code with a unified code base.
You need back-out plans. You need to identify all risks, rank them, mitigate them, build contingencies, and ensure that anything you screw up can not only be undone, but identified. Messing up data accounting in invisible ways will tend to run for a long time before anyone notices the error. This means parallel deployment as a test environment--possibly for months--to validate your outputs on real-world workloads against the existing system before deploying the rewrite.
That's only some of the bare, early technical pieces. That doesn't include procurements, human resources, buy-in, communications plans for literally everything that impacts anyone, management of legal and regulatory concerns, and the like. This isn't a small project; it's a massive effort to remove decades of technical debt and reduce long-term operational risk. It's like the removal of a cancer that's wound its way dangerously through tissue, but is still not fatal and can be cut out without destroying the body: the longer it grows, the harder it's going to be to remove, and the more likely it is to lead to sudden vital organ failure. Banks and insurers are functionally-immortal and don't simply have to outlive their barely-viable systems, like a human who might make 90 or so and die of heart failure before the prostate cancer even starts making them ill; they can actually keep running forever, and die when the system fails.
Your COBOL program is probably running under CICS. The hardware CICS and your program have been running on have been optimized for your workload. Your data is probably stored on ECKD DASDs. Your data is probably stored in packed decimal format, so it can be operated on with a single, optimized, machine instruction.
I've seen data-crunching mainframes in insurance companies running modern code. Not C#.NET, Java, and Python; they've got JCL and Rexx to chew through massive data sets and run huge reports. They use mainframes because the software on the mainframe and the I/O hardware does in twenty minutes the job a $300,000 bank of servers can't do in literally three weeks. They still use software written this century, and they've left off software written back when mainframes were "The Future". That software doesn't look like anything today's Web developers are familiar with--no PHP, no C#, no MongoDB-
When I was on amphetamine, I started using soylent because I could only get 1200kcal/day. The original dose (20mg MAS ER) was so powerful I couldn't eat at all: food had a terrible taste and texture, and swallowing felt like an invasion. It was hard to get down. I eventually settled on 15mg MAS ER; 10mg gave me anxiety, and 15mg only made me mildly-depressed. I break over at 20mg like an avalanche effect, complete with a mix of mild and moderate-severe overdose symptoms. The pharmacological window for amphetamine is supposed to be huge, like 8:1, so a large dose usually makes you kind of high and causes insomnia (any dose causes insomnia; it takes 26 hours to be capable of sleep after taking amphetamine), while very large doses do slow damage and get you really, really high (consequence: you're going to feel shitty later and, eventually, you'll reach a point where you're never going to feel good again).
By contrast, alcohol doesn't seem to work too well on me. Benzodiapezine-likes (Eszopiclone, a non-benzodiapezine) seem to not do much at normal doses. Enough alcohol will get me drunk, but it won't muck about with my reasoning abilities; the whole experience is terrible, involves a mixed manic episode, and leaves me incapable of sleep as if I'm on a stimulant--even at 0.34. One of the annoying things about the emotional clusterfuck is the emotional clusterfuck, which I'm stuck observing because of course I am; you'd think I'd just go along with the ride, but no, emotions are an external factor and not a driving part of my experience of self, and I'm prone to point out where emotions are eroding my judgment and overrule them (largely because I'm not comfortable with the emotional decision--figure that out). Eszopiclone puts me in a daze, during which I notice I feel really fucking high 24/7, and that my memory is iffy; it doesn't help me sleep and it does nothing until several hours after I've taken the pill.
Drugs suck. Somebody get me some better tools. In your case, you need... something. MSG in your food, or a good dose of Welbutrin (...no thanks) or Phenylpiracetam (I like this stuff; it's relatively-safe). PPR makes my reward system work correctly and allows me to experience feelings of emotional pleasure, but I generally don't use it, and I remove it from my schedule whenever I have actual doctors making drug adjustments so I've got a better baseline during treatment. Even when I do use it, I take a lowered dose to essentially activate wanting without activating the emotional side of liking, because feelings of pleasure still register to my senses as being really, really high and I'm not entirely comfortable with the impact on my judgment (which is stupid: lack of reinforcement behavior has destroyed my judgment, and my primary goal is to avoid any and all discomfort, so I'm not really bothered by failure but don't pursue achievement if it means expending effort; I need good emotional feelings, it's a requirement for proper function, even if it's scary and kind of embarrassing to take drugs to make me feel good).
Ask your psychiatrist to fix that shit. Trust me, being broken like that sucks. The alternative requires just as much discipline, and draws less sympathy--people will criticize you more for getting fat than starving yourself to death--and it's still a lot better to have to push yourself into eating less than you burn and maybe monitoring your intake and caloric consumption instead of just stuffing your face on the couch. When your body isn't telling you food is good, life becomes difficult and miserable.
19x19, 13x13, or 9x9. The number of possible positions is 2^361, but the number of possible moves is a lot bigger. The problem is projecting those moves is long and difficult; you have to be able to extrapolate localized influence and identify how that impacts a global strategy in an abstract sense, or else you can't play. A computer can't track all of the possible moves because losing a big position might be less-relevant than a play elsewhere that establishes power to restrict your gains to 1/3 of the board and solidify the entirety of the other 2/3 for the opponent.
The amount of analysis is huge. You can create algorithms to shorten it, right up until a human does something strategically-novel.
It's the poor who are most-impacted. The general middle-class are less-impacted, and some are frequent fliers; the upper-class may or may not be frequent fliers, and don't spend all their time on airlines. The very-rich have private jets. Everyone at every level tries to optimize for their particular case, but may have a less-expensive option (e.g. 1.5% cash back is worth 1.5%; 2.2% frequent flier miles are worth 2.2%).
As well, the upper class are fewer and gain less by spending on consumer goods. They buy more in luxuries, and besides have limited purchasing: even if Warren Buffet shops at Giant with his AirMiles card, he's only buying Warren Buffet's groceries; there are a ton of people with lower-fee cards shopping at Giant, and so the average is impacted almost not-at-all. Meanwhile, Elon Musk might buy some fantastically-expensive diamonds at a jeweler on his card for some hot girlfriend of his, which might have some sort of impact on the price of diamond necklaces poor people can't afford.
Most of Elon's money isn't going through his personal credit card; it's going through his business ventures. Warren Buffet's money is mostly in brokerage transactions, buying and selling securities, so he's essentially using his money to suck your money out of your 401(k) (yeah, you put your money in funds without being as good a trader as Buffet, this is what happens; you should have had a plan, or just bought the 3% guaranteed interest option instead of dreaming of possible 14% returns on high-risk bonds in the "Aggressive Growth" fund). The money to pay these people still flows up from the bottom; the impact isn't catastrophic.
It becomes a vague system when you realize it's non-homogeneous and still works that way. The practical impact is essentially that the group of people shopping at the same merchants transfers money between themselves based on that distribution, which means the cost of stuff at the Save-a-Lot is probably not terribly impacted in the sense that the poor are bringing poor-people spending habits and poor-people rewards (which they also pay for) and getting little to no advantage/disadvantage (a wash). The stuff at the upper-end jeweler whose cheapest ring costs $80,000 is probably more broadly-distributed across various tiers of rewards--and the jeweler probably only accepts credit because rich folks expect to have credit cards accepted, and will pay the mark-up for merchant fees (you try to buy a motorcycle, the dealer will let you put $2,000 down on a credit card--the rest must be cash)--and it's all rich people one-upping each other on who has the best rewards structure. The common intersection is mostly middle-class and poor, and most of us think we're getting something (mostly, we're evading the penalty for participating by getting a 4.5% discount on gasoline, 2.8% discount on groceries, and 1.5% cash-back on everything else--which we pay for whether we collect or not).
It's almost-correct to say that the rewards programs essentially cost nearly nothing to anyone, except that there are people who don't participate optimally and there's still a spread between people whose optimal behavior brings higher versus lower merchant fees to the merchants involved. In truth, it's partitioned in odd ways, and participation engages multiple partitions to varying degrees, resulting in optimal behaviors which are not necessarily optimal in monetary return. Basically, you don't have a choice how it impacts you, aside from whether or not you play along to minimize the cost to yourself.
The business owners were prepared for it. Somebody changed the rules. Now they have to have a $250,000 back-up tractor that's going to sit around not doing a damned thing, rusting out, and becoming obsolete. They won't be able to use it once it's obsolete because they can't replace a part that breaks; since that part's not made and John Deere won't repair the tractor anymore, they need to junk it out. Farmers want a 20% profit margin, but the market only gives them an average of 10%, sometimes 8%; that means that one piece of equipment represents profit off $2,500,000 of revenue, and some farms are quite small.
What do you do if a change in business operations at a supplier can have that kind of impact against an operation? What does CareFirst Blue Cross do if the failure of a RAID drive in a server results in the server shutting down for 2 weeks until a Dell technician comes out with a hard drive, replaces it, and resets the RAID controller so it knows it's okay to rebuild the array and turn the server back on? Note that, today, a failed disk in a RAID array doesn't cause a service interruption, and Dell will have replacement hardware in your hands in under four hours; an immediate halt of service and a two-week lag time to get a new drive installed would be a major change to business operations.
What do you do if shit like that happens?
We spent $160,000 here over a few months to analyze the results of an upcoming event. We have some really smart people who determined who would buy what and how much they'd pay, and projected the most-likely outcomes so that we could position marketing and take actions which would cause outcomes worth several tens of millions versus just entering blind using a glance at historical data and some expert judgment.
Do you honestly think banks don't do that kind of simulation? It's short-term, multi-variable statistics calculating the probability ranges of a market, not long-term economic models trying to figure out what businesses or sectors will be strong or how the global market will grow across the next 20 years. Banks don't buy this shit in bundles and lose 40%; they purchase what their customers accrue, likely on a quarterly purchase, carrying over what's left from last quarter.
Customers might handle things less-well, and they pay for the privilege. On the other hand, I'm pretty sure that frequent fliers are actually getting frequent flier miles; they might not be using them in the most-efficient manner, but they're getting them and using them.
Sleep restriction. Take your total sleep time, set your bedtime by rolling back from your desired rise time. When your sleep efficiency over a week reaches 90% or more, move bed time back by 15 minutes. If it falls below 80%, move it forward.
I'm an insomniac with a prior ADHD diagnosis. I got Modafinil from a psychiatrist after my attention issues became asinine; it was fucking awesome for 2 weeks, then I got hit with sudden suicide-grade depression. Modafinil doesn't interrupt my sleep; and since I was sleeping 2 hours a night for over a year anyway, the consequences of abusing Modafinil to stay up for weeks on end suddenly became clear to me--which is actually great, because I wasn't able to find that in the literature. I also got a sample of suicidal impulse, but not much; the experience of that kind of extreme depression was too-new and carried too much risk, so I shut those thoughts down as they formed. I'm too hardened now for that to happen again, so I won't get a second chance to examine how the actual desire to kill yourself forms in a safe environment. Oh well. I get why people bitch about depression so much now; I learned to make it go away, but it was never that bad, holy shit.
That lead to a huge run of psychiatric experimentation. Amphetamine for the anhedonia? Makes me depressed (yes, it does the opposite of what makes people want to smoke meth). Melatonin? No dice. Wanted to avoid a GABA drug, so my psychiatrist gave me Belsomra (Suvorexant), which worked amazingly: instead of 2-3 hour sleep latency and lots of being awake for hours at a time in between, I got to sleep in 20 minutes, despite not actually feeling any sleep urge--I needed like 20mg of Melatonin to actually sleep, though, even with the Suvorexant. $300/month so insurance wanted me to try GABA drugs, so I tried Eszopiclone; it made me really fucking high for a couple weeks, then I stopped taking it because I nearly drove into another car 20 hours after the last dose; the withdrawal was hilariously bad, but lasts only a day.
I'm on Atomoxetine now. The drug hit me like a hammer, and 80mg/day was making me manic, driving my heart rate up by like 30 points, and causing ludicrously-high blood pressure. 60mg still hits me with fatigue, so I want to split the dose and cut it back more. Since I started taking Atomoxetine, I can control my insomnia with little more than sleep restriction, although 2mg of 8-hour continuous-release melatonin helps, and an overdose on B6 helps immensely (B-100 supplements have a toxic level of B6; B6 toxicity only shows up in literature after a minimum of 1.9 years of supplementation, mean 2.6 years, and can cause permanent psychiatric damage. Most "dietary supplement" sleep drugs--unregulated in the United States--contain high doses of B6, e.g. any Melatonin pill will contain 5x the RDI, and ZMA contains a lot of B6. ZMA is also essentially a GABA drug like Valium, since Magnesium binds to the Benzodiapezine site on the GABA[a] receptor and induces a similar anxiolytic effect).
So yeah, it's Atomoxetine and sleep restriction, at a minimum, for me. It's fantastic. Meanwhile I see idiots on Dimaxion and Everyman thinking they're performing better because they have less sleep, and I'm like... I remember I didn't realize the sleep deprivation was actually causing me a problem, even though I knew it was miserable. My brain started shutting down major facilities to compensate near the end.
I've considered that Ramelteon might make my life a bit easier, but I don't want to introduce a new dependency into my life right now. I'm curious, because it hasn't been directly-compared to Melatonin in literature yet. (Ramelteon has a 1-2.6 hour half-life; I get no sleep urge. Sometimes, I'd just like to feel like sleeping when it's time to sleep, instead of feeling like I'm dying of sleep deprivation but still not feeling tired and having sleep be just a damnable chore. Imagine if you only had to eat because you'd become dizzy and weak eventually, and food wasn't really that palatable; eating would be terrible, but necessary.)
It's a tractor. You sit on it and ride. It's not a stationary machine or a commercial jetliner.
We're not talking about an extra $10k; we're talking about the tractor sitting around non-working while enough time passes to call that $10k per minute because it leads to a several-percent drop in yields on a million-acre farm. It's not as direct in farming as it is anywhere else: if you get back on in X time, no big deal; if you have to wait 2 weeks, everything goes to shit. An afternoon of annoyance and some overtime is a lot different than losing a dozen days out of a 100-day growing season.
It's a good strategy, and carries different and more-immediate risks. It's an attractive option for me because it allows factoring in new lessons-learned and it makes risks controllable: old COBOL code might not respond well to new requirements, and will eventually need to interface with new code--either new COBOL code or new code in some other language. You can't control those risks as well as you can control an elective rewrite, because the risk of new requirements is often surprising and under time-pressure, while an elective rewrite to remove technical debt is at your best pace unless one of those risks suddenly appears--in which case, you may decide the rewrite isn't ready, and that shoring up the existing mess a bit longer is less-risky.
An elective rewrite requires accepting those risks now and devoting the resources to the project immediately. That translates to "take an active step in causing disruptions and expenses that we don't need to", versus "take an inactive step in causing disruptions and expenses at a future time when we can't avoid it". This form of proposal requires reframing.
Reframing is a known method for improving decision making and reducing dorsolateral prefrontal cortex activation when making decisions. Generally, to make a difficult decision, the dorsolateral prefrontal cortex must activate on prefrontal reasoning and override the mid-brain response (the simple, automatic response) by force of expended energy. For a frame, "Do you want $5 today, or $10 in 6 weeks," you want $5 right now; it takes a lot of effort to resist the immediate reward. A null-framed form, "Do you want $5 today and $0 in 6 weeks, or $0 today and $10 in 6 weeks," produces a decision to wait for the $10 in nearly an additional 40% of the respondent pool. Essentially, the prefrontal cortex sees these proposals differently, and the midbrain has a reaction to both explicit conditions rather than a lack of response to the implied null conditions--and obviously finds getting $0 at any time offensive versus the alternate, so is somewhat less-focused on time as a factor.
So do you want to take controlled, manageable risks now and pay an expense to avoid uncontrolled, time-pressured risks in the future; or do you want to do nothing now and face uncontrolled, time-pressured risks at various points in the future when implementing a long-term solution to avoid more such risks suddenly cropping up afterwards will not be immediately viable?
An event which will occur at a random, probabilistic period and incur some penalty (call it $1,000,000) is less-desirable than an event which will occur at a random, probabilistic period and incur some smaller penalty (call it $100,000). The cost to replace these events--to mitigate future risks--can be high (e.g. $20,000,000); yet those future risks may be more variable ($10,000-$100,000,000, mean $1,000,000) versus the alternative ($15,000-$250,000, mean $100,000). Frequency may also be a factor, wherein a new system may already have the flexibility for many upcoming new requirements, and any changes may bring with them opportunity to cover additional requirements without gold-plating (e.g. it has to handle PNG images, and you used a library that handles 47 types of graphics formats--when it has to handle TIFF embedded in EPS, it already does that, and nobody went out of their way to spend additional time making that happen). Making the system ready for commodity environments means we can build test environments and more-readily detect defects before production changes, as well, without expending millions of dollars supporting said environments.
It's not a simple problem. It requires a reframe to present it as a complex but important consideration, rather than a scary hobby to keep someone busy. After that, you're ready to sit down and analyze the problem in its full scope to decide if you should proceed. I've had times when the answer was no, and that's usually when we're migrating to a new system soon an
The cardholder doesn't pay; everyone pays. The cardholder with the highest-merchant-fee card is not paying that merchant fee; it's averaged across all transactions, including cash (0% fee) transactions. That means cash holders pay the largest mark-up and get nothing; while high-merchant-fee rewards cards with no cardholder fees pay the same mark-up (read: no additional cost over just paying cash) and get the maximum return (read: what's purchased by those merchant fees is paid for mostly by all other consumers).
Everyone pays into the system. The guy with the best rewards comes out ahead; the guy with cash comes out behind; and the guy closest to the average-fee rewards card essentially gets a wash.
Actually it would seem the extra merchant fees are diffused through merchant pricing, as the customer doesn't pay extra to use his card. For example: if fees of 1.5%, 3.5%, and 2.2% show up, averaging 1.8% of all purchases (including cash transactions with no fee), then all goods sold by the merchant get marked-up by 1.8% to include those costs. That 1.8% pays for the frequent flier miles (3.5% fee on those cards).
Those miles are purchased with consumer after-tax money, and taxing them as additional income would be double-taxation of consumer income. QED. By the by, the 1.5% cash back you get on your card is also paid for by ... well, the rest of us who have lesser cards with lower rewards, or those paranoid schizophrenics who only use cash to keep their movements hidden from the gub'mint.
Frequent fliers are, by definition, flying frequently. The summary seems to suggest that frequent flier miles are a huge free hand-out to airlines; yet aren't the frequent fliers actually consuming those miles, rather than letting them default and thus paying for them (or, well, diffusing the price through consumable goods, thus having everyone pitch in a little to pay for them) and not using them?
Upstart and traditional SysVinit pretty much ignored output; if your script didn't send things to syslog or custom logs, then it went nowhere. Pretty much the first thing I noticed about systemd was how much less-frequently I had to modify init scripts to figure out what was wrong.
Mostly my dislike of Mir was because of its NIH bullshit that I got tired of with Redhat already. Wayland shows up, Canonical is like "well we'll write our own!"
Pretty much, Canonical claimed Wayland is inadequate in many ways and Mir does things a whole lot better; a bunch of people (including the Xubuntu developers) evaluated running on Mir, and found it shrugworthy. We've all basically decided Wayland is where everything is going, when everything does eventually get around to getting off X11, and so Canonical is wasting programmer time that could go into improving Wayland instead.
As an example: Wayland has an XWayland client compatibility layer and a weston-rdp compositor. Wayland can seamlessly transition to different compositors, meaning it can draw on your graphics card for a while and then swap over to weston-rdp. XRDP required an entire Xorg server compiled as an RDP backend, and is defunct. It's possible to muck about with XRDP to get Sessman to launch weston-rdp with XWayland; with a bit more effort than that, you can get a display manager running on tcp/3389, and allow selecting (and displaying) log-in sessions over RDP. Basically nobody cares, and instead we have people busy A) just trying to get Wayland up to production-quality; and B) making Mir because Wayland was not invented here. More people could work on problem (A) or (C) if they skipped problem (B).
With Unity going away and Mir probably vanishing with it, that's years of programmer time lost to defunct garbage that never gained any traction.
"Google has heard of it"
The software breaks every time you make a change. Need to pull out the transmission and replace a damaged gear fork? The software now requires an authorized technician sign off on the repair, and disables the tractor.
The problem is it cost the farmer $500 to do the repair on the previous model themselves, in an hour. The new model also costs the farmer $500 to do the repair themselves, in an hour; however, the tractor won't "go" until they pay a technician an additional $1,500 to drive out, wave their badge at the device, and whisper the secret word into its ear.
Part of the complaint is they can't fix their tractor and get back to work; they have to take a relatively-significant hit to productivity and put their farm at risk waiting for a service call.
He's just making shit up.
So global statistics show population growth is slowing even though we're still growing population faster than ever because population won't really stop growing until it stops growing?
History has shown, and continues to show, that advances in technology reduce scarcity, and that reduction in scarcity directly causes a population boom. Reduction in scarcity is wealth. It's the capacity to feed 7 billion people on a planet that can sustain 0.63 billion humans. It's the capacity to have cars, roads, and rail lines when medieval society couldn't build all that shit in 100 years even if given the industrial population we have. It's the capacity to afford a wardrobe of clothes--you know, rich people things, since people who aren't the Nobility can't produce enough to make it economically viable for most people to own more than one or two sets of good clothes (which eventually get torn up and used as rags, because throwing things like that away is ludicrous).
This is what you're seeing throughout history.
We used to have a revolving door, but I'm working with coworkers who have mostly been here for 3-5 years now. They're all cognizant of the market; we regularly talk about how we can't hire additional people because we pay way below market rates--which includes our pay, considering some of us have been offered 80% pay increases at other jobs doing the same shit.
Some of us are stable. Others just don't care to deal with a job change. A few thought the job would be interesting and now are afraid of jumping ship without 2-3 years to show they're not job-hopping. Plenty are placated by the continuous change in our environment and have a perception that things are getting better, never mind that they're not getting paid more. We're all roped in for various psychological reasons that are quite clear to me.
I'm working at a business with low pay--where the average for a programmer is $96k here, programmers make $74k. The same is true of most IT staff, running a good 20%-30% short of the industry median.
We're also fairly diversified and have chicks and people from all over the world in our staff, and have had folks who speak Russian or obscure Indian dialects as a primary language in prominent technical positions. They're also poorly-paid, although near as I can tell we all have about the same salary.
It seems like a form of posturing: we don't want to pay salaries, so we create a perception of ... something. We're a good place to work because of something something benefits diversity open-door-policy.
Are these studies by industry, region, experience, and business? Do we say that black women earn 55% as much as white men, or do we say that black women at business X in job Y earn 55% as much as white men in business X at job Y? What happens if business X mostly hires white men for job Y, and business X' hires a higher proportion of black and asian women for job Y but also pays like shit even if you're a white man?
That works after you've put in the time to learn to cook, which involves some planning. Most people aren't thinking that far ahead.
So they Johndeere'd the iPhone for a passable reason?
The researchers in lab coats are making big money. A lot of it is their wages. The wages of people at chem synth companies, mining facilities, and other supply chain services is also a big part of the majority cost.
The question is: what fraction of the cost did that overhead reflect? I've seen gross-mismanagement, and I've seen stuff that's so damned expensive that a 0.5% extra expense is... a lot.
Executive salaries are typically interesting in that respect: the actual cash compensation for most executives amounts to pennies per hour. The CEO of Ford gets, from revenue, about 1.05 cents per employee per hour, or $21 per year; stock and options compensation come at the expense of diluting stock--basically the business pays nothing, issues new stock, and investors in the secondary security market all wind up holding pink slips they can sell for slightly-less than they could yesterday. That means Ford basically robs a penny here and there from your 401(k) to pay its CEO, CFO, etc.
People see executives with $20 million compensation and talk about how they should give that to their employees; then you find that their cash compensation (salary and bonuses) is only $7 million (most of which covers taxes on the other $13 million), and their 300,000 employees would each receive $23/year or a raise of 1.17 cents per hour.
So are the R&D perks 40% of the budget or 0.4% of the budget?