Slashdot Mirror
Ask Slashdot: Has Your Team Ever Succumbed To Hype Driven Development? (daftcode.pl)
marekkirejczyk, the VP of Engineering at development shop Daftcode, shares a warning about hype-driven development:
Someone reads a blog post, it's trending on Twitter, and we just came back from a conference where there was a great talk about it. Soon after, the team starts using this new shiny technology (or software architecture design paradigm), but instead of going faster (as promised) and building a better product, they get into trouble. They slow down, get demotivated, have problems delivering the next working version to production.
Describing behind-schedule teams that "just need a few more days to sort it all out," he blames all the hype surrounding React.js, microservices, NoSQL, and that "Test-Driven Development Is Dead" blog post by Ruby on Rails creator David Heinemeier Hansson. ("The list goes on and on... The root of all evil seems to be social media.") Does all this sound familiar to any Slashdot readers? Has your team ever succumbed to hype-driven development?
Describing behind-schedule teams that "just need a few more days to sort it all out," he blames all the hype surrounding React.js, microservices, NoSQL, and that "Test-Driven Development Is Dead" blog post by Ruby on Rails creator David Heinemeier Hansson. ("The list goes on and on... The root of all evil seems to be social media.") Does all this sound familiar to any Slashdot readers? Has your team ever succumbed to hype-driven development?
I think infinite web pages was the worst idea that every site just had to copy to be part of the fad. I liked page number buttons. I can bookmark a page where I left off instead of scrolling a hundred times from the top again. It also doesn't use up all my computer's memory in Firefox or Chrome.
Main issue isn't "following the hype" -- it's not understanding why something worked for someone, or even why what you're currently doing is or isn't working before making sweeping changes.
PHBs making stupid and declarations based on trade magazines that sinks the project? Probably never understood what his subordinates were actually doing in the first place.
Developers changing languages mid-project? Forgot to add the time to master the language to the estimate, most likely.
Although, it was due to a sustained level of hype, rather than an epiphany by the powers that be.
Sig ?
It happens all the time. Everywhere. As soon as a trend arises on the horizon many companies jump on it to get their share of the cake. And it's even not unprofitable.
Slashdot, fix the reply notifications... You won't get away with it...
The problem is that experience can do one of two things to developers. Open your mind or close your mind. Many programmers refuse to open the Pandora's box and they stick to a tool, paradigm or coding style they know even though its not the best thing to solve the problem at hand. It's like a carpenter trying to cut down a tree with a circular saw because that's what he spends 99% of his time using.
Wow, ExtJS brought all development to a complete multi year halt. In the first few months ExtJS development is way way way faster than any other framework out there. But after about 6 months all you are doing is fighting with the framework. Just an endless knifefight. Any single problem could be solved against the base instlall of ExtJS but what happens is that you have to develop workaround after workaround to make the system snap into place for any given need. Those workarounds then make future "easy" changes impossibly hard.
So you might have something as simple as wanting to put the focus on a login username. If you had just done the page as your first round and thought of that then, like everything with ExtJS, a little weird but fairly easy. If you already have fought with sencha to make other things happen on the login page (say a filtered twitter feed) then ExtJS is probably broken 8 ways from sunday and you can't set a focus worth a damn.
Save yourself a world of pain and just use basic javascript combined with either simple single function libraries, or worst case scenario use a framework that won't blow up your company like react or polymer. Yes, you won't be a showoff in the first few weeks of development like you could with ExtJS, but you won't blow up your company when you can't finish the project until you realize that it can only be done by throwing out ExtJS.And if you get 5 or 6 people in the company who get training by ExtJS, good luck cutting through her bullshit about how ExtJS is the best thing ever even though the project is now 18 months late.
Many programmers refuse to open the Pandora's box and they stick to a tool, paradigm or coding style they know even though its not the best thing to solve the problem at hand.
Precisely the OP's point.
That's a typical trait of a junior developer, or an experienced developer who has worked solo for most of they're career.
For some projects and some teams, Agile is the best they can be expected to do. For other types of projects and other types of teams, it's a really horrible idea.
Central to Agile is the proposition that the company is unable or unwilling to figure out what the requirements are before they develop the system. As Yogi Beara said, "if you don't know where you're going, you not get there." On small projects it might not hurt too much to figure it out as you go along, to backtrack and throw away code that has to be replaced. On large projects, and systems that need to integrate with other systems, you REALLY do need to figure out the requirements ahead of time and plan the architecture.
If your team consists solely of programmers of medium competence, Agile may be the best choice. If you have even one excellent systems architect, you're far better off letting therm do their job, planning the system out first. If your team includes junior programmers (or veterans who haven't expanded their skill set over the years), Agile can leave them floundering, going one direction for a few weeks, then another direction for a few weeks, then completely backtracking for a few weeks.
In summary, Agile is sometimes the best choice for your team, and when it is, you've done a poor job of hiring.
Several years ago my Pointy-haired Boss was reading technology articles (bad idea) and caught the "Big Data" bug. It spread to our CTO, CIO and all department heads like wildfire. This led to our Development team being turned into NoSQL zombies who said words like "Hadoop", "Shark", "Spark" in response to any new product requirement. It was a glorious vision of a magical backend system that would take all our data from every platform, that would scale up and out forever, and could be asked any question and give us exactly the results we wanted all instantly. The fact no one in the entire company had ever used any of the technology before or the fact we didn't even have any Java experience to setup even the base Hadoop installation were just minor points not worth discussing. I would like to say I was the lone dissenting voice, well I was and said lets just stick to SQL, but even I got caught up in the hype eventually.
18 months later and a sickening amount of man-years wasted and contractor money spent with no usable products or services the conclusion was NoSQL isn't a good fit for our data or platform use case. So they all went back to standard MySQL and completed 90% of the delayed projects in under 4 weeks.
On the plus side management heads did roll. I have a new My Pointy-haired Boss and CTO. However they have now started to drop in the words "Microservices" and "Docker" into all discussions. I can see a new hype-train arriving shortly ...
I used to work for a (bigger than small but smaller than medium) family owned business doing web development.
The CEO and President were brothers, they were the sons of the owner.
One of the two was the idea man. He'd see something on a competitor's website or he'd read about it somewhere and call a meeting to find out what would be needed for us to do it too. We'd discus it, start developing a plan and get to it and three days later, there'd be another newer, shinier thing that he wanted us to work on. It was soul-crushing because we never got to follow through on anything. I was very happy to leave that place.
LK
"Hi. This is my friend, Jack Shit, and you don't know him." - Lord Kano
The problem is that people want magic solutions, and they keep chasing the latest fad in the hopes of finding the secret alchemy that will make average developers turn into gold stars, produce perfect systems in a tenth the time, and meet all requirements without the bother of knowing them.
Anyone who's ever done any system of significance that actually worked will know that the "best" tools and methods are situational. Need a bash script to list a few files? The approach is different than it would be if you're hired to redo everything used by the IRS.
We can go all the way back to the "shelf full of binders" methodologies. In their day, they were supposed to be the magic cure-all. Today, it's Agile, or it's XYZZY or whatever is the latest and greatest. Still haven't found that secret sauce.
One size doesn't fit all. There is no magic. Successful development projects require skill, experience, good judgment, hard work, and competent leadership.
Central to Agile is the proposition that the company is unable or unwilling to figure out what the requirements are before they develop the system.
The problem is that waterfall is presented as making extreme effort to try figuring it all out up front, while Agile then becomes to the exact opposite where you make no effort and just prioritize what's right in front of your nose. Reality is that you need some flexibility in waterfall projects and some structure in agile projects. In my opinion it's fine as a development method, it's all the people making requirements who don't even try anymore because agile. We're so dynamic, as long as we can spin in place it doesn't matter that we're not going anywhere.
Live today, because you never know what tomorrow brings
Once upon a time I worked on an app that had 4 databases - MySQL, Redis, Neo4J and Influx. Each of these were to solve a specific problem (searching, time-series data, etc) even though the scale of the application (a handful of users per day) never warranted any kind of "big data" solution. And the fundamental problem remained - many of the developers didn't know how to write decent SQL.
Postgres / HSTORE could have probably solved pretty much the entire set of persistence use cases. But that's a solid, proven and ultimately boring technology. Where's the fun in that?
It's not just PHB driving the madness. Plenty of it comes from resume-driven development.
Here are ten easy steps that you can take to implement Hype-Driven Development for your project.
1. First, choose a new tool. Find somewhere that the tool is being used by a company everyone has heard of. Don't be too concerned about what they're using it for, or whether it relates to your work in any way.
2. When you start using the tool, don't mention it to anyone until you've already decided to base your finished product on it.
3. Don't bother finding out if the tools you have can already do the job you're doing now.
4. Expensive tools are automatically better than cheap tools. This makes it easy to measure fitness-for-purpose.
5. Even if you only use the tool to simplify very mildly half a line of code that's only used once, incorporating a new tool is still worth it.
6. Compare the tool by re-implementing some of your existing tasks. Only test the simplest and most trivial scenarios: if it works in a simple case, it's bound to work just as easily in a complex case.
7. Any inconsistencies with existing standards can be readily overcome by creating a new standard that the new tool fits exactly. Try not to be disheartened by the idea that you've previously been doing everything wrong for years.
8. Have some like-minded suckers re-implement everything even vaguely related to the new tool from the ground up. The more suddenly you can implement this, the more of an impact it will have - and impact is always cool.
9. If the re-implemented product turns out to be awful, or if it doesn't do what users want or need it to do, you'll be committed to the new tool by then, so it won't matter. Tell anyone who is critical of the product that it's too late to change it and that they should have raised their concerns earlier - especially if they did.
10. Stride confidently into your next performance review, knowing that even though you wasted a lot of time and resources to build a product that does slightly less than it used to, you've certainly achieved a lot.
Attack its weak point for massive damage!
...has proven itself for five years. The hard part is convincing executives of the five year rule. Often the benefits only appear in narrow niches or under specific conditions, but it takes a while for the industry to learn when and where.
Also, a lot "fads" are not directly technology fads, but rather obsessions. About 2 years ago our CIO became obsessed with SEO - Search Engine Optimization (Google hits, more or less), and so all kinds of silly games were played with our Internet content and CMS's, including mass repetition.
After a while people realized there was too much content to manage and clean up. That CIO moved on and the new CIO is a minimalist. Big change. SEO did nothing but make a mess.
We were suspicious of it all, but there was nothing we could do at the time but go with the flow. At least bullshit = jobs.
Table-ized A.I.
> There's nothing preventing you from running an agile project with a robust and complete design.
A large project with a complete design, an actual plan, may be agile (the adjective), but it's very much not Agile (the development methodology). A core tenet of Agile is that design, planning ahead to the end of a project, is impossible. In fairness, it probably IS impossible, for the people who believe that.
If they haven't been taught one particular trick, they probably never will be able to know the requirements before they write the code - trial and error really is the only option, if nobody ever told you the method to find out the real requirements.
If you want to know what the actual requirements are, there's one way to find out (and maybe ONLY one way). Sit down with the user and watch them work. Ask questions as needed to understand their workflow while they actually do it, and take notes. Ask the actual user, not their manager's manager, what they need to do their actual daily tasks. That way, (and probably only that way), your User Stories aren't fictional stories imagined by some manager, they are real descriptions of real users doing real work. Requirements flow directly from there.
I disagree.
The experienced developer has been chopping down trees for years with an axe. He's been putting up shelves with a drill. he's been cutting floorboards with a circular saw. And occasionally he's been cursing because he's having to make do with the wrong tool because although he knows what the right tool is, paying $lots for a tool he will use just once can't be justified.
Alongside that there are countless (less experienced) developers suggesting that he uses the circular saw to cut down the tree, the axe to put up the shelf, the drill to cut the floorboard and the experienced developer isn't particularly impressed.
But in the back of his mind he's always got that thought "what if that next tool is the chainsaw. " Just think how many trees I could cut down then. But even when the chainsaw comes along, he continues to use the circular saw on the floorboards, the drill for the shelves and, indeed, he may even still use the axe from time to time.
God said, "div D = rho, div B = 0, curl E = -@B/@t, curl H = J + @D/@t," and there was light.
I agree. I cannot wait for that fad to die an ugly painful death. Make the pages longer, that's fine. But not infinity.
I hear it causes ADA lawsuits. I hope so, sue 'em hard!
Similar annoyance points for the "flat" look. You cannot even tell a button is a button, and entry box boundaries are washed out. Shade the fsckers, people! It's not 1989.
Table-ized A.I.
You ask on stackoverflow, of course. *eyeroll*
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
It depends on if it really matters. A little script to pop up a box and let the user choose from a list of environment variables before running an application could be done properly with python or it could be done in a couple of lines using zenity in the bash shell. Sometimes using that circular saw is pretty damn quick if the tree is small enough that it doesn't matter.
That's why planning is important and choosing the technique before taking a major step is the way to get that experience. Solving toy problems or making small changes can be done with just about anything so that's not the way to start using the right tool for the job.
Reading Slashdot comments it seems that many seasoned developers are dismissive of some pretty good new tech, even after it's been around for much longer than 5 years.
C# is a great example. I'm a hard core C coder who mostly works on embedded systems, but when I need to do anything desktop I always consider C#. It might not be the most efficient language, but it's performance is perfectly adequate for a huge number of tasks, it has libraries that simplify most day-to-day stuff greatly and lets you concentrate on structure and architecture instead of details.
People around here often dismiss it because of the association with .NET (which itself is far from terrible) and the fact that it hides a lot of the "real CS" stuff, but that's the point of it.
Save the hate for stuff that deserves it, like Javascript.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
I did a contract with Sun, just before they went under. The employees were quick to tell stories about how they used to hire magicians to come and entertain on the campus. There was this one guy, sat a cube over from me. Near as I could tell, his job was to sit on the phone all day boasting about whatever next conference he was going to and how he was a certified black belt. That was the only time I'd ever heard anyone talking about it. At Sun. Just before they went under.
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
Wow, ExtJS brought all development to a complete multi year halt. In the first few months ExtJS development is way way way faster than any other framework out there. But after about 6 months all you are doing is fighting with the framework
This is the good old "Rapid Application Development" myth that has been doing the rounds since before many of today's trendy agile NoSQL programmers and the PHBs who encourage them were born. Even with things like Microsoft Foundation Classes and Borland's OWL that were used to create substantial apps, the initial drag-n-drool honeymoon didn't last very long. Then when Multimedia came along there were more "authoring systems" than applications created using authoring systems, and they were all great until you hit the brick wall that the designers had never anticipated, and ended up re-writing from scratch.
"Ooh look, I can create a fully-functioning GUI app by clicking 3 buttons and writing 1 line of code... this is going to save weeks of development"
Six months later: "Ooh, look - I'm wading through the sparsely-commented source code of the framework trying to figure out why I can't get the 'print' method to do anything beyond the trivial case given in the sample project... what's this? '/// TODO - implement print function'...?" (Its too long ago for me to remember the details so I won't name the application framework in question).
Turns out that a couple of days writing the "boilerplate" code for your application paid dividends further down the line...
First example I remember was this: The Last One - yes folks, the last computer program that would ever need to be written, heralded by magazine articles predicting mass unemployment of programmers... but I'll bet you an internet that Ada Lovelace had some brilliant ideas for making the Analytical Engine take all the hard work out of programming...
Put simply: we all know that the last 10% of development takes 90% of the time. RAD tools eliminate the first 10% of the development thus ensuring that the last 10% takes 100% of the time.
In a survey of 100 programmers, 111111 thought that duck-typing was a good idea.
People around here hate C# (those that do) because it's from MS. When it comes to MS, there are no technical merits that can redeem the technology. They are not rational people. Most of them probably don't even program for a living.
FWIW I often compile with Mono because I like my tools to be cross-platform. I find that compatibility between .NET and Mono is excellent anyway.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Yes it's as good as you would Expect and is the proper way to do an involved script GUI interface.
However sometimes a couple of lines of bash using zenity does the job and is pretty obvious to anyone who has to alter it later.
Exactly! So when waterfall tries to nail down the details up front, it does itself a disservice. Users find out months later that those nailed-down details don't quite work for them, and now you have to go back and re-engineer your project. With Agile, you find out much earlier in the process that the details weren't right.
There is a reason to hate C#. It is effectively controlled by a single corporation. Don't give us bullshit about it is "standard" or "open". It isn't. It is Microsoft. It doesn't matter if it is Microsoft or Google or Sun or Apple or whatever, it is still controlled by a single corporation. C++ is open and standard.
People that don't like C# (like me) don't like it for a number of reasons starting with lock-in, sub-standard libraries if you're wanting to be cross-platform, etc. C# was a reworked clone of Java after MS-Java was found infringing. The CLR is interesting, but is a fundamentally different solution than the VM approach used by the JVM. It solves the many languages running on Windows problem, not the run language X consistently across multiple architectures and OS problem. Hence the lock-in issue, because if I'm going to run on Linux, I'll use Java, not C#, as C# offers me nothing compelling to use it on various flavors of Linux, BSD, OSX, OS400, etc. And no, Mono isn't good enough and EF (ORM) isn't a reason either.
The cesspool just got a check and balance.
I dislike C#. I have programmed with it, C++, and C as well, even relatively recently. For my particular purposes, C/C++ wound up being far better languages to code the system I needed (on Server 2012) than C#. I needed system calls that required calls into the Win32 subsystem directly, and if I need to write a library in C anyways and call it via a lightweight PInvoke wrapper, why not just write the entire thing in C and skip the extra complexity, overhead, and debugging headaches?
The cesspool just got a check and balance.
People around here hate C# (those that do) because it's from MS. When it comes to MS, there are no technical merits that can redeem the technology. They are not rational people.
The complaints I've heard didn't generally sound so irrational. I thought the consensus was "It seems like a good language, but still most useful in building things for Windows. Maybe that will change as the cross-platform stuff improves, but for now, I'll stick with [whatever language they're using]." Admittedly, I'm not a real programmer and only get a sense for what programmers think from this site.
Too often developers choose to use a technology because it will look good on their resumes, not because it serves the interests of the system's users or the people paying for it. It's what economists call "agency costs".
Every time a new golden hammer comes up, developers rush to use it before they get left behind. And you can see the corrupted focus right in the code. I remember when Model-View-Controller was the buzzword du jour, and people without any sense of irony whatsoever would bake MVC framework dependencies into practically every single file. Ugh.
But here's the rub: part of taking care of user and customer needs is considering the impact of future brainshare. Sure, COBOL may be just perfect for this app (OK, probably not), but should you really saddle them with having to find someone with COBOL expertise? It's possible to be too puritanical about avoiding technology fads.
So part of your job as a developer is to track the emergence of new golden hammers, to study good and bad examples of their use, and to truly understand each of them as much as possible. Where possible you should try the latest thing; if you're a team manager assign slack personnel to do spikes that evaluate it (this is a great perk to hand out). It's part of your job to stay on top of where things are going, without committing customers to something that might not meet their needs.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
I'm a perl programmer, almost by definition I don't get hired by places that insist on chasing the new shiney.
The tendency of programmers in general to be as trendy as a bunch of teenagers has not been lost on me, however (like I said, I'm a perl programmer).
Somewhat more disturbing is a tendency of perl-culture in general to be a bit faddish... one year it's inside-out objects, the next year it's the Moose family, one year Module::Build is the greatest, the next Extutils::Makemaker has made a comeback and no one wants to hear about anything else, one year ORM are the bees-knees, the next it's the NoSQL fad, then it suddenly dawns on people you don't really want to try to do schemaless data...