Developer Experience: Fundamentally harder than normal UX
gabrielpickard.comTo make some of the ideas in this article a little more concrete, here are some research demos I’ve made:
* Legible Mathematics, an essay about the UI design of understandable arithmetic: http://glench.com/LegibleMathematics/
* FuzzySet: interactive documentation of a JS library, which has helped fix real bugs: http://glench.github.io/fuzzyset.js/ui/
* Flowsheets V2: a prototype programming environment where you see real data as you program instead of imagining it in your head: https://www.youtube.com/watch?v=y1Ca5czOY7Q
* REPLugger: a live REPL + debugger designed for getting immediate feedback when working in large programs: https://www.youtube.com/watch?v=F8p5bj01UWk
* Marilyn Maloney: an interactive explanation of a program designed so that even children could easily understand how it works: http://glench.com/MarilynMaloney/
These are very interesting! The Legible Mathematics one also made me think of APL, which is very hard to read - but specifically it's hard to read for the same reason that traditional linear programming math is hard to read, the lack of visual grouping[1]. I wonder how it might be possible to redesign something like APL to take advantage of rich text and more complex formatting?
[1] The weird symbols might seem like an initial barrier to this, but they're only hard to read if you're unfamiliar with APL; they're actually very easy to learn and remember, as there's not that many of them and they all do very basic operations. However, APL loves operator overloading and likes to give operators different functions based on whether they're used in prefix ("monadic") or infix ("dyadic") forms, and there are also higher-order operators that consume the operators immediately adjacent to them and then operate on the expressions after that; all of this makes the nominally right-to-left parsing require a fair bit of mental effort instead of being able to rely on immediate visual recognition.
Great experiments/demonstrations.
What are your thoughts on Mathematica/Wolfram Language? Some of these ideas are present in it (like mathematical typesetting, interactive documentation and live code/data updates).
I don’t know much about it. From what I’ve seen it looks like something Stephen Wolfram can use really well but I wonder if other people are able to use it.
I think it has its merit. It is a very _designed_ system which is definitely a side-effect of Stephen Wolfram's attention to it. But despite it being a bit of a singular vision, I do think that it presents a very complete idea of what a programming/computer interaction environment can be. For example it is like Lisp in that it is homoiconic and the whole interface, layout/formatting, the file format and everything builds upon this.
I really love your Flowsheets prototype, I hope someone pursues that direction in the future. “Spreadsheets with more tools from traditional programming” seems like such a powerful idea.
This looks fantastic! What a treasure trove.
> We've simply gotten used to them: Dealing with the idiosyncracies of bash, vi, or the JavaScript type system
This stuck out to me, there seems to be a trend in UX/UI where any move away from the "simplest path" is seen as a huge negative. Could it be the case that we use these tools (especially UI patterns like vi) because after the learning curve the give a huge amount of value? It seems like we are assuming that we should make a developer tool with the same level of "immediate familliarity" that we try to build into a website where customers will bounce easily, for an audience who is willing to spend time learning a tool if it provides value to them.
Rich Hickey has a pretty interesting take on this very topic where he compares programming languages to instruments, and how instruments aren’t for beginners. How it takes work to become good at playing an instrument, and how that isn’t a bad thing.
> But look at this guitar player with blisters. A harpist has blisters, a base player with blisters. There's this barrier to overcome for every musician. Imagine if you downloaded something from GitHub and it gave you blisters. Right? The horrors!
That whole talk is filled with some interesting takes on designing and building software (with the usual skew that paints Clojure in a good light, so take it with a grain of salt if necessary).
[1] https://github.com/matthiasn/talk-transcripts/blob/master/Hi...
> Imagine if you downloaded something from GitHub and it gave you blisters. Right? The horrors!
I'll have to remember that next time someone mentions that C should be a dead language.
(I actually think C is a fine language, but should be deftly handled)
> Imagine if you downloaded something from GitHub and it gave you blisters
You mean Haskell?
I'm a designer [0] and an engineer — you'll take my shell from my cold, dead hands.
There are two issues here:
a) Designers trying to simplify everything beyond usefulness is a good instinct gone haywire. Simplification helps, but without an understanding of accidental complexity versus essential complexity, one is bound to end up painted into a corner with no flexibility left in the app. Few designers understand this, and those who do got it the long way round — by working on products that have a lot of essential complexity, like AdWords, and by repeatedly fighting those battles
b) An engineer's operating environment, OS, IDE, shell, terminal, is a reflection of the inside of his or her mind writ large. Like every Jedi has to build their own lightsaber, every engineer has to go through this pain of building out their weapon, because one's workflow is how one thinks, how you look at the problems at hand. No UI designer can help with that.
[0] (Because it's relevant to the context: ex-Google, ex-Facebook as professional experience)
I like this idea of how "one's workflow is how one thinks".
There is a great little book - Daily Rituals [0] - that goes into many artists' and scientists' daily habits. The habits are very much along the same lines as your thought - they are workflows for how the individual tends to think best.
I'd love to see someone put together a website or book that did that in the context of software engineers' workflows. Does someone know if a resource like that already exists?
[0] https://www.goodreads.com/book/show/15799151-daily-rituals
-edited for grammar.
I digress but I am really interested in your experience as both dev and design as I am going in this path too.
How did you juggle acquiring skills for both roles? Also, where there times you had to struggle between the two mindsets on a task?
>An engineer's operating environment, OS, IDE, shell, terminal, is a reflection of the inside of his or her mind writ large.
Shout out to the people that don't really bother because they're good enough to adapt to any setup. :)
Look I can adapt just fine. Even onto projects that run on Windows if I really need to, but investing time into making your personal environment better and faster is really, really worth it.
Shout out to the people that use the best tools for the job because they're good enough to optimize their setup. :)
For things like vi, I'm developing the "left handed mouse" analogy:
Many mice are ambidextrous (e.g. the Apple puck). Most are weakly right-handed with a slightly assymetrical shape. Some are very strongly right-handed (e.g. vertical mice) and can't be used sensibly in the left hand. So left-handed mice also exist.
Some people are naturally left-handed. We (as a civilisation) used to treat this as aberrant but have now recognised it, and that different tools suit different people.
I believe that something similar exists in programming tooling in relation to how people think about programs. There are clearly some people who have a strong, unusual "handedness" and have developed tools to match (e.g. Colorforth). A few people discover these and find them amazingly usable. Most other people find them baffling.
Consider the three propositions:
a) Jimi Hendrix played guitar in the wrong way with the strings in the wrong positions
b) Jimi's configuration was correct and everyone else was wrong, because he's producing the objectively best music
c) Jimi was left handed, and had constructed an accomodation which worked for him but should not be expected to work for anyone else
Far too many discussions of programming tools devolve into (a) versus (b), largely because people want there to be an objective ranking of who the best programmer is and what the best tools are, rather than allowing for diversity of (programmer x tool).
> aberrant
Pretty sinister, if you ask me.
Maybe a bit unrelated, I got used to Windows 2000, then window XP, then Gnome 2, then Gnome 3 came. So I stuck with Gnome 2, then moved to XFCE, and now with RHEL 8 I had to use Gnome 3 because there were no other options. Gnome 3 is an absolute horror show. I don't know who made it, for who, and what the theory behind it is but I don't see how it would be easier to use for someone non technical, my parents both understood how to use Windows 2000. It is just weird. It is harder to multitask as efficiently as I did in XFCE and Gnome 2. Gnome 3 is not simpler - it is just more convoluted.
And I feel this is a very similar situation with other tools. I edit code with vim, in a terminal. This is simple as dirt. I do it because it is simple as dirt. Visual Studio is incredibly complicated to me because to do the creating code part my job I need to understand the following:
- How code is built.
- How to build the code without using any graphical front end.
- But now when you bring VS into the mix I need to also understand visual studio. It does not remove complexity, it adds it.
Similar thing with debugging, I need to understand all the ins and outs of debugging but now bring VS into the mix and I need to understand it's stupid UI.
I like simple, my mind is simple. I can learn things, if there are rules and patterns it makes it easier to learn, but the less things I have to learn the happier I am. I don't have an option to not learn some things, like how to do build automation, how to debug code, how computers work, etc. But I do have an option to not learn something entirely useless like VS.
I think the lie being sold is that somehow you can be a programmer without actually knowing how to use a computer. And to know how to use a computer is not the same thing as knowing how to click on things in the UI with a mouse. To know how to use a computer you need to understand how to use it to do automation - and once you need to do this VS is just a nuisance.
Just a rant I guess.
I think you are missing an important aspect: VS can help you in mental tasks the same way a jackhammer can help more than a pickaxe in physical tasks.
Sure, the pickaxe is simpler, but it will take you a lot longer to break that concrete with a pickaxe than with a jackhammer, because of simple human limitations.
Similarly, refactoring and advanced code navigation support takes a while to learn how to use, but once you do, it empowers you through technology to quickly do things that your mind would take much longer to do by hand.
For example, say you want to extract some code from the middle of a function into a separate function. With vim, you would generally have to manually move the code, write the new function header, then start to painstakingly inspect the code to identify the parameters to pass between the two. You will probably make mistakes and have to wait for the compiler to find them. All in all, assuming it's a bit of hairy code, it might easily take you more than an hour to get it working. In VS, you would type ctrl-r, ctrl-m (Refractor, extract Method) and it would automatically detect all of these for you, pop out a dialog box, you'd enter the new function name, Tab, param name, tab, param name etc, and enter when you're done. Maybe 5 minutes all in all, assuming you also do some ctrl-r, ctrl-p (refractor, parameter) afterwards to extract some larger expressions back into the original function.
Similarly, you have things like 'analyze data flow to...' which can find all places in your program where a particular value can be written to, and do that recursively until you get to the original source. Same thing - this can be done by hand with a series of finds and so on, but an advanced tool will just help you do it faster.
But, just like with advanced editing in vim, you need to take the time to learn the tool until you can get the most out of it. Same as the first time you enter vim you're likely to fumble to even be able to exit, you can't expect to be productive in VS if you don't take the time to learn what it can do for you.
VS is great for visual debugging, the build was secondary: you set it once and forget. But it only worked in the era of the proprietary software because you didn't have much third-party stuff to link into your project. Testing has changed too: you'll write an automatic test anyway, which means that the software is shaped for debugging by writing tests.
VS is decent at debugging if it does not fall apart in the process - which for me happens more often than not. So not really useful.
The only way I found of making Gnome bearable these days:
I also use it, but it is quite buggy and not quite enough. Luckily xfce is now available for RHEL8 via EPEL.
> then moved to XFCE, and now with RHEL 8 I had to use Gnome 3 because there were no other options.
Wait, what happened to XFCE? I use it right now.
RHEL8 doesn't let you install a different desktop environment?
I'd look into switching to Fedora. I use the XFCE spin and it is stable.
I think there should be a clear distinction between your tools and... let's call them "utilities" what comes to UI. It is perfectly okay to make electricity plug, water faucet, toaster and the power button on your computer not only so easy that an idiot can use them, but so easy that an idiot can't use them wrong. Not only because requiring mental energy to use these is irritating but also because it can be dangerous.
Your tools, then again, chainsaw, microscope, text editor... have no reason whatsoever to have an UI that is intuitive without training[1]. Because without training you are anyway going to be either dangerous, useless or in best case just really unproductive.
[1] of course, the UI needs to be efficient after training.
And, in the case of things like chainsaws, they shouldn't be needlessly dangerous if a user is not as well trained as they should be or if they're simply inattentive/tired/etc. (And in the case of software, you should generally not be able to cause massive damage/loss of work because you picked the wrong menu item.)
My point of view is different: there has to be a really strong reason to move my eyes from the code, and in the editors there should be respect for that. For example, I don't want to move from Atom to VSCode because VSCode has those huge icons that take space and distract me, when all I want in the default editor UI is the project directory tree and the code. Everything else should be opt-in.
For what it's worth, you can easily hide those icons and access everything through the menu or command bar.
Right but VSCode doesn't work well with that bar disabled: for example, now I hid the activity bar (that's how it's called it seems), I clicked on a .status file and VSCode said "There are extensions for this file". I clicked on that, and now I have an extensions bar at the directories tree bar place. This happens because this UI setting is not the default one, so you encounter such kind of confusing behaviors every time you move away from the default behavior.
Is that a problem? There's only one sidebar. I agree there could be more, but I don't think there's much to be gained by it. Just learn the keyboard shortcut to bring up the folder tree. Or customise the icons so that they stand out less. Heck, I keep the sidebar collapsed most of the time since it's so easy to toggle it on with the keyboard.
I'm talking about UI/UX, not about shortcuts. If you switch my important sidebar with a marginal one it's a problem to me, from the UI/UX point of view. If you really want to talk about shortcuts: I don't want to learn the shortcut for the directories tree, because I just never hide it. And I'm talking about my flow, which might be different from yours.
It actually does work well with the bar disabled. Everything is emacs accessible , the menu is just a wrapper around text commands, like they are in xemacs.
I've given an example of what doesn't work when you move away from the default behavior, that's the point
In "The Design of Everyday Things" Don Norman describes this distinction as knowledge encoded in the head versus knowledge encoded in the world.
For professional tools knowledge encoded in the head supported by appropriately encoded knowledge in the world absolutely is a viable approach, provided there's appropriate feedback and conceptual mapping corresponds to the mental model a user has about how that tool works, i.e. actions and reactions should be consistent.
With modal design patterns such as the ones used by vi, for example, this can become a problem.
Listing vi in there shows that the OP, even if they may have good points in general, is limiting what good UX is to certain specific attributes, whereas developers want to optimize for additional attributes.
And this isn't just me saying that because I like vim. It's because of the objective fact that nearly every developer tool that is created will include a vim mode. And if included as an extension it will often be one of the most popular extensions. What that objectively indicates is that there is a large contingent of developers who genuinely find the vim modal editing UX excellent, to the point they seek it in other tools as well (including browsers, mail clients, RSS readers, etc).
Yes, you are right I didn't put that very well. Modal editing really has a lot of benefits, even if I personally don't go for them (am more of an Emacs guy).
What I would say is that vi really doesn't give you a lot of interactive context -- and it's hard to add it on.
I think even complex products should have a linear progression from newcomers, curiously exploring the product for the first time, to experts, who want to do certain tasks as fast as possible.
It's simply not possible to learn vi by just using vi.
Also, the author emphasizes that just dumbening a product is not the solution.
Any examples of software that accomplishs this well?
A quick summary of my experience:
- Most complex apps are not learnable linearly: Ableton Live, Final Cut Pro, Logic Pro X, Photoshop, After Effects, Blender, all practically require reading the manual.
- A few apps are ok at it: Adobe Lightroom, and Sketch, although those apps also are probably less powerful than those in the first category, e.g., Photoshop can do the majority of what both of those apps can do, and more.
I would actually put shells and text editors as some of the easiest complex apps to learn linearly, because you can do so much with them with just cut and pasting text from the internet. Try following along with a Blender tutorial video that's not for an absolute beginner, and you'll get stuck almost immediately, because you won't know the keyboard shortcuts to perform the actions in the video. This happens far less with programming tutorials involving text editors and terminals.
> It's simply not possible to learn vi by just using vi.
I did. I learnt vi when I was told go fix this file on that computer, you have ssh, x forwarding won't work because you have 3 hops (embedded devices that won't allow forwarding of any kind), there is only vi on the box.
So I figured out how to use vi. It is not rocket science, there is not that much to learn, vi muscle memory takes time but also not even that long I think.
I certainly learned vi from vi (or I guess vim to be specific). The start page says "type :help or press <F1>", which led me to the help pages (and maybe doing vimtutor, can't quite remember), and then I "learned vi(m)" just working at it, poking and prodding at the various corners (and the manual for is excellent). No Google, no stack overflow, no teacher. Later I started looking up other peoples configurations and stuff, as well as some advanced guides, but I was already basically proficient at that point.
On the other hand: I don't see what's so bad about a product for professionals where you might need a teacher to help you learn. Photoshop is not a simple product to use if you've never interacted with it before and most users go through some tutorial or has a teacher or something. It doesn't mean that Photoshop has bad UX. This goes for lots of software made for professional use: try learning Autodesk Maya or Avid without external resources, I dare you.
Did you read about vi before? How do you know all the shortcuts?
I googled it. I did not know vi before that, and now I use vim as my main editor for code.
So you did not just "learn vi by just using vi". You have also used a browser, google, and tutorials found by google. There is nothing wrong with that. In fact congratulation on learning vi! But sadly your example does not illustrate that one can "learn vi by just using vi".
The intro screen tells you exactly how to open help and exit. Man gives you even clearer instructions.
If you just launch vim it tells you the basics of how to use it, and how to get more help. So it's definitely possible to learn it without external resources.
Emacs is the same.
There's not much to learn if you're editing the odd file ... if you want to use vim as an effective development tool, there's loads to learn.
I would think there is a lot less to learn with vi than with Visual Studio.
Oh yes, that is certainly what we do. And there's nothing wrong about that. We are definitely in the business of building tools for professionals, i.e. a bit of a learning curve is not the issue. The "professional hazing process" isn't necessarily bad. What I would say is that there is some (maybe a lot of) potential value that vi cannot develop simply due to the very reduced form factor that it has. I believe this never really has come to the fore because most programming languages are designed in a fairly limited way. I.e. they don't really take into account that they are a User Interface.
As positive examples I would point out the kind of interactive editing mode that you can find in dependently typed programming languages. I believe e.g. Idris has a pretty cool Emacs mode.
Bash and JS are interesting cases because they're so ossified core problems are impossible to touch.
Far from me to defend every UI of every development tool out there but I think statements like these, and their illustration, could benefit from some explanation for those of us who are not designers:
> We coders still put up with horrid UX/UI when programming.
which is illustrated with a screenshot from Visual Studio... .NET 2002, I think, judging by the application icon?
Setting aside the relevance of a 20-year old screenshot, what exactly is wrong with that interface and what makes it horrid? I mean it definitely had its quirks but:
- It's spectacularly compact, certainly way better than anything I've seen in the last five years. We could display an UI builder and the associated code on single 1024x768 screen and work on it semi-comfortably. "Beautiful" UI/UX, as understood today, is so cluttered by whitespace (oh, the irony...) that it's barely usable on a 1920x1080 screen. A similarly compact interface on today's huge screens would be a dramatic productivity improvement that, twenty years ago, we could only dream of.
- You could easily access any function through textual menus -- no hamburger menus, no obscure, monochrome icons. Granted, the toolbar icons were a pain, but the way I remember it, most of us either disabled it straight away, or just populated with a couple of items that were of real value and which we knew well.
- The colors have great contrast, the whole thing is readable even on a very poor-quality screenshot that seems to have been actually downsized.
- UI items have enough relief and/or distinction that it's clear what you can interact with and what you can't (maybe the item palette from the diagram editor is an exception, or at least the screenshot makes it look like one, but virtually every program in that era made it look like that so it wasn't so hard to use).
So what's wrong with that thing?
I think of most UX designers as children: they want everything to conform to the latest trends and are always in search of the new shiny thing to play with.
Having worked with designers since back when Visual Studio .NET 2002 was pretty fresh, I think this is a very over-simplified assessment and I don't think it's fair.
There are definitely cargo cult designers, just like there are cargo cult programmers out there. But I don't think that's true of most designers, just like I don't think it's true of most programmers.
Chasing the latest trend isn't always a decision you make by yourself, and isn't always a decision you can oppose on your own terms.
Virtually every designer who was in the industry ten years ago or so can come up with a good design that's 100% against the latest trends -- contrasting, non-flat, compact, whatever.
But it'll get shot down within minutes in any product design meeting. Sometimes by people who have zero design experience, so they can't judge a design except by how well it conforms to the latest trends. Sometimes by people who lack the "political" capital to argue for an original design with their bosses, too. There are a lot of factors here, and most of the time the people who do the designs have the least amount of influence.
The problem with developer tools is not technical, it's all about the selling part. If you make a new superior developer experience, you think developers will instantly see the benefits? haha! You first have to teach them, then after a few month if you are lucky you will get a "aha, now I understand". So first you need to manually educate each user until you have a critical mass. Then you need to market and hype your product. So that developers will tell each other how cool your new technology is. Continue with that a few years until there are code in production that use your product, before even thinking about a business plan. So there are few options, either you have enough money so that you do not have to "work" again, and can spend your time making new tools. Or your current employer lets you work on the tools. For a startup working on developer experience (language and tools) I would suggest a 10 year runway (funding) and that 1/3 of the budget goes into educating users and 1/3 goes to marketing.
Any software business that is profitable, already have code in production, and that code need to be maintained. So instead of creating a new better experience, you can make the current experience better. eg. putting rockets on a horse, rather then creating an automobile.
> The problem with developer tools is not technical, it's all about the selling part. If you make a new superior developer experience...
Selling is going to be hard but I feel like you underestimate the technical difficulty of replacing a large stack of complex tools that have decades of work and experience behind them. And that, in part, makes selling harder: I'm immediately suspicious of anyone who claims they've invented a superior way to work. It's more likely that they've invented a small improvement (and an arguable one at that) for a particular scenario, but developers would still have to rely on their old tools for a lot of stuff. In worst case, they're trying to sell a tool that doesn't extend but replaces the old tools without providing support for scenarios and workflows that existed with the old tools; step forward on one front, three steps back on others.
Of course, small improvements to existing workflows can usually be implemented by developers for themselves (and others while at it) once they learn about the idea, and that's how the developer experience has slowly improved over the years.
For example, you can make a new fancy code editor (let's call it sublime) and hype it on features like multiple cursors. And I can have that in emacs at the cost of about 3000 sloc of elisp, and I don't have to give up any of the old things that I've grown to rely on.
But this is exactly why Rails did so well. The developer experience was good, right off the bat, in a way that a 15 minute video could portray so that developers could instantly see the benefits. There was no "manually educate" step taking months.
You are so right, those "build a website in 15 minutes" demos hit like a bomb back in the day!
I told my UX buddy half a dozen years ago that I was optimistic that there seemed to be a glut of UX people coming because we were going to steal a bunch of them to look at DevEx issues.
I spent some time nerding out over woodworking hand tools a few years back and it pretty well cemented for me something that I’ve suspected for most of my career: people down in the muck have very limited vision. That your output is only better than your input by degrees.
I’m not sure there would be much fine woodworking at all if the best woodworking tools were only as good as the best software tools. There is no Lee Valley of developer tools. You can’t make me stop using JetBrains (individual licenses were their best idea every), but it still doesn’t rate above a Woodriver, and if I’m honest some of their stuff is Stanley level, and not even the antique stuff. And their stuff is better than just about any other tool I use all day.
I suspect Harbor Freight could make better software than Atlassian, and I don’t even mean that as a metaphor. I think I could take harbor freight employees and get better requirements out of them because they wouldn’t be up to their ears in cognitive dissonance.
Atlassian truly is the Harbor Freight of developer tools!
I'm a bit hopeful that DX might finally start getting the love it deserves. -- For better or worse, Microsoft seems to understand the potential that lies in building better tools and seducing programmers to join their fold.
Disclosure: I used to work for a competitor. Some of my coworkers were afraid of Atlassian and it rubbed off on me.
Later when I used their stuff and could be (more) objective, I was sort of confused by that previous experience. They're... okay. If that was intimidating maybe that said more about us than about them.
I am probably experiencing a little backlash when I react to their shenanigans because they are not 'all that'. At all.
Trac (no affiliation) is ugly but like your favorite hammer. It is probably the least painful project management suite I ever recall using. And I used it before they decoupled their parsing and rendering logic, so writing plugins that involved anchor tags was a warren of duplicated code.
My last musing-while-brushing-my-teeth was to wonder what would happen if someone drop-kicked the Trac CSS files and wrote a new one from scratch using modern conventions.
I don't think that DX is harder than UX because developers somehow have a more complex task than everybody else, and that there's therefore a richer "experience" to navigate than in other domains. I think it's because developers tend to be developers, so are more aware of and willing to accept the nuances and complexities that go into development. In other words, we've got a more detailed understanding of the developer experience because, as developers, it's what we experience.
We accept the heritage of developer tools - the keyboard-driven interface that displays to a teletype emulator, the edit-compile-debug workflow - and build tools that improve the processes that have built around those legacies. This is why when something comes out of left field like Adele Goldberg and colleagues describing Smalltalk, we find it easy to adopt the approach to code organisation on offer and hard to adopt the image model, browser-based workflow, debugger-driven iteration, and other changes.
Meanwhile, when we go out into other domains, we use a little bit of understanding of that domain, a lot of reasoning by analogy, and an intention to "disrupt" what already exists and "eat the world", and create something that works very well for the spherical user in a vacuum without all of the detailed understanding that comes from having grown up in the system and learnt from people who grew up in it even longer ago.
A healthy exercise, I think, is to replace "developer" with "scribe" whenever we have these conversations. It becomes clear that in large part it is the overall culture that is missing the point -- which should be a kind of mass literacy. Smalltalk is a different universe indeed, but its goals were also completely different from today's. Its creators assumed that "using" a personal computer would be, in part, "programming" it, and they sought to define what that interaction would be like.
We need to find our equivalent of a printing press! For what it's worth, possibly that won't be just one thing, but a wide array of tools?
> For what it's worth, possibly that won't be just one thing, but a wide array of tools?
We need to think hard about what reading and writing really are. These are the only technologies of their kind and they have had a completely different kind of qualitative effect on individuals and whole societies than other types of media and/or technical systems.
It brings up interesting questions. If you have a scribal culture, what are the interests of the community of scribes? Would they support easier writing systems, more amenable to learning by outsiders, or something more arcane that protects their cultural and social position (or something in between)?
Also what is the transition from one literate state to another like, what does it entail, how long does it take, and what kinds of "works" appear in the interregnum? In the Middle Ages the Universities in the west were largely concerned with endlessly rehashing Aristotle (a thinker on the border between orality and literacy) over and over, without coming up with original ideas. Is the enclosed community of developers today more like the scholastics than they'd care to admit?
There are stark differences between systems like Unix and those hinted at by Hypercard, Oberon, Smalltalk and the rest. It's like we know about the alphabet and what it is capable of, but the cultural inertia of complex cuneiform systems is too much to get past.
Isn’t that a circular feedback that leads nowhere? Developer like to be developers but being one doesn’t mean to make solutions and profits. Like a bureaucrat likes bureaucracy and then everything is bureaucracy under the hood sprayed with marketing sauce. Best stories of best products sound “we hacked, broke, short-cutted, ignored, made a fucking lots of money, and then rewrote fundamental instruments to match our needs”. Isn’t there a controversy with a nice cool environment that developers praise?
> There has been tremendous growth in the field of UX/UI. Without a doubt, today's applications are much more user-friendly than those from 30 years ago. Back then, users were given features, interface be damned.
That 30 year time frame takes us back to 1990 and back then the user experience was limited by the technology of the time.
However a decade later we had Windows XP.
I would say that 20 year old Windows XP might in fact be a much better user experience than the modern UX/UI we have to live with today.
The much less powered CPUs of that time felt much more responsive than the modern day CPUs/OS that we have to today.
> How do I run this thing? Where does the code start? Is my system configured correctly?
This is why there are occasional spasms of "back to basics" or plaintive remembering of the BBC Micro. You power it on, it beeps, and within a second you're in the interactive development environment. Typing code runs it directly. Typing code with a line number adds it to the program. No configuration, containers, downloads, updates, dependencies or uninformed choices to make.
> Why do we treat this as a moral failing instead of a usability issue?
Yes. This applies in so many places. Learn from "Poka-yoke". The system should make it easier to do safe things and harder to do unsafe things.
> Tests are a usability dead end
Depends what you mean by "tests". A strong type system does away with certain categories of test (and conversely a lot of the heavy unit testing usage comes from communities with weakly typechecked languages). But both types and tests are capturing a human-level requirement of "if X then Y", a constraining of the problem space.
This is why many successful code archaeology maintenance projects start by building a test suite to capture the current functionality of the program. An executable requirements document.
> Tests are a usability dead end -- I know this may be contentious, but I believe test suites are another realm of excessive moralizing en lieu of better tools and better processes. Too often they function as a security blanket that simply encases the parts of the code that are unit testable, while leaving the vulnerable, untestable bits fluttering in the wind. Approaches such as generative testing seem more promising.
I'm not quite sure what the author is arguing for here in particular.
Anything that makes writing tests a bit easier, e.g. suggestions for additional test cases, would be cool, but ultimately tests are about writing down your assumptions/expectations about the code.
No, they are not formal proofs and sometimes they are not perfect but they still provide a lot of value. So far I haven't found a good reason not to write tests (since I outgrew my newcomer attitude) and yeah integration tests are usually what I focus on most. For any case where testing whole systems today is hard, there are some fundamental challenges (e.g. end to end web UI test). I don't quite see how tooling will get rid of the need for tests.
> So far I haven't found a good reason not to write tests...I don't quite see how tooling will get rid of the need for tests.
I write device control software. It’s very difficult to have true automated testing of things like drivers. You can write unit tests for subsystems, like packet parsers, but integration testing generally requires good ol’ “monkey testing.”
“Just write a mock!” Is what I hear all the time.
Mocking a device is a massive project; potentially larger than designing the device, itself. Remember that the mock needs to be of unimpeachable quality, and also needs to do things like simulate adverse signal environments.
DX for that kind of thing can be awful.
As far as basic DX goes...
Most developer tools are wrappers for command-line OS tools, and it shows.
They can also be quite buggy, and we accept this bugginess. I use Xcode, which is quite “crashy.” I am constantly fixing issues by deleting the build folder.
Back to testing...
I prefer test harnesses over unit tests. I write about that here: https://medium.com/chrismarshallny/testing-harness-vs-unit-4...
Just a question. Do you imply that the mock of a device should simulate/emulate the entire behavior of the device? Because you definitely don't need to do that and it would, of course, be a vast and hard-to-justify endeavor.
You only need to _mock_ the object, that is emulate behavior on a small scale.
Yes and no.
It depends on what we are testing.
It can actually be worse than useless to rely on the "perfect world" of unit testing. I have had a nasty shock, when switching from mocked behavior to real-world behavior. I write about that here: https://medium.com/chrismarshallny/concrete-galoshes-a5798a5...
I'm not going to argue against your experience and from reading the article it seemed like you made the right call. Perfect world testing is indeed a failure, but I would say that "imperfect" world testing is exactly where mocking is beneficial, because it can reliably produce all sorts of anormal behavior. But again, it obviously depends on the situation and circumstances, and good design probably removes more errors than good testing.
Sorry, I didn't mean to imply everything can be well tested, but where it can tests make sense even if they don't cover everything (vs. the attitude that testing is pointless unless you can cover all relevant code paths).
Also, I don't think there's as much value in unit tests than other more holistic types of testing, i.e. functional/integration tests.
TBH I read your linked article and still don't quite know what a test harness (in that context) is and in what way it's better and differs exactly in the concrete. Guess I'll have to read up on it elsewhere.
I appreciate the feedback. I suppose that I should probably add a paragraph that describes what a test harness is.
Something along these lines:
"A test harness is a [usually] simple standalone application that uses the CuT, and presents a user interface. Since it is for testing, it may expose internal metrics and allow the user to access parts of the code that are usually hidden."
This is very true. Not coincidentally, driver flakiness is an ongoing problem, as is the tendency of OEMs to bundle the actual drivers with a huge ball of mud (NVIDIA, HP are major offenders).
Perhaps some sort of formal spec for the hardware communication API would help, and would also help on the ASIC side, but I can't see how such a thing would be built and popularised. It's a very small balkanised world of driver writing.
Too often, "tests are required" -- but as you say it's very hard to test anything meaningful, so developers "test something" and call it a day.
I think that improving DX might be the most important thing of the future software industry. If a single developer devotes his live to a better DX, hundred of thousands of developers might be significantly more productive.
I really like to explore new DX approaches (just recently, I published an extension for VS Code enabling visual debugging [1]). But I find it hard to make a living out of it, as so many companies find it granted that everything is free. They would rather hire another developer than paying for licenses that might effectively increase the effiencency of the developers they already have.
[1] https://github.com/hediet/vscode-debug-visualizer/blob/maste...
Just a little bit of feedback, while the visuals are quite nice I think this page could be helpful if treated like a landing page, what's the problem/why is this helpful compared to what I currently/normally do. I might not be the target audience though, since I don't understand what it does/why it helps quite quickly through scanning
What a cool extension! I'll give it a whirl!
On impact: I've been running the numbers for a tool I'm working on and the projected savings for the industry look insane! Just by shaving off 10 minutes here or there you can contribute a lot.
This. Despite latest <libname>-cli trend, anything slightly more sophisticated requires to create a structure, then create structures in nodes, then a configuration structure, structures all the way down. One could object that everything we do is structures, but there is no ui to these, only a text file and a formal documentation (if you’re lucky).
Text files and their disconnection from a documentation is a root of all our evils. Not only they diverge with new versions of everything, but there is a constant attention switch (stacks of them!) and unnecessary diving into things that may or may not be important to the development process. There is no way to omit these checks when you learn or return to an idle project.
I have a long time idea that every config, format, api call, and so on should come with inseparable documentation ui (+rationale, examples of use, best practice links, pre-configuration, diff/merge views, etc). Yes texts are simple and easy to read, but we also write. You can make text from a structure in O(1), but you cannot make a knowledge from an empty file in O(sensible), for any sensible sensible.
Not arguing on salaries though. Even pretenders who have no clue can take a great cut off this nonsense.
> but you cannot make a knowledge from an empty file in O(sensible), for any sensible sensible.
why isn't a schema and then auto-complete a sensible way to do config?
Both xml and json has a schema file that you can create generic text editors with auto-complete for.
Try this:
I taught Scratch to kids. They love the visual element of the language. It's so hard to make basic mistakes - partly because you can't put an Int where the code expects a Bool.
The interface shows you all the components you can use, and what data they require.
I then move on to teaching Python, and watch kids get frustrated.
Languages like DRAKON should be the future of our profession - not typing 80 char lines into a terminal.
The question is: how do you grow up from there, in terms of expressive power? I.e. is it possible to find more middle ground between Python and Scratch?
The other day I was using MusicBrainz Picard and it occurred to me just how absolutely pleasant it is. It uses standard desktop widgets, is fast, has some nice attention to detail and consistently accurately relates the state of things, is powerful, and stays out of your way. It—and this is #1-with-a-bullet more important than every other UX concern—behaves consistently.
I don’t think it would survive a pass from most “UX” folks in such a nice state. It 1000% wouldn’t survive a designer or hybrid designer/UX person (it wouldn’t look pretty in screenshots on their portfolio).
The main problems in software tools are lack of consistent behavior, lies, and tons of ways to use a bunch of tools that all do basically the same thing (and you’ll probably have to know more than one). The hardest part’s not using them, exactly, it’s knowing all the different, stupid reasons they break. It’s a general quality issue more than a broader UX thing, I think. That extends to libraries. And I don’t also mean tools and libs from big names—I mostly mean them.
A site about UX that uses a fixed-sized font with horrible spacing? This I about the worst website UX I can imagine... Long live reader view...
Haha, I agree. Sorry.
I can't be the only one who prefers the way applications looked 30 years ago? When they dumped whatever they had to offer right there for you to explore and pick from. These days everything is hidden below a million layers, trying so hard "not to be in your way" that you need a deep love of pixel-hunt-point-and-click adventures to get anything done..
Let's take an example: Specifying DNS servers in Windows. 95: 1. Rightclick "Network" on desktop -> Properties 2. Doubleclick on the TCP/IP protocol for the NIC. 3. Type new DNS server. 4. Press OK.
10: 1 Press start 2 Search for control panel 3 Open Network and Internet settings 4 Select Change Adapter Settings. 5 Rightclick NIC and select Properties 6 Select TCP/IP and then Properties. 7 Type new Server 8 Press OK
Sure, some may think that Windows 10 looks better than Windows 3.11 or 95, but I don't and I can't believe everyone does.
I think developers are missing the testing practices UX designers have grown accustomed to:
1. Usability tests where developers literally sit down and watch someone install and use your library from scratch (I find a lot of developers do not like to do this). Things like, seeing where they have to look up documentation (and how they do it), what bugs they hit, and how often they make common mistakes. I think a lot of this could be logged e.g. A developer signs up, you have their email + API key, you can connect the dots between what doc pages they view, how often, and what errors they commonly run into.
2. Doing whatever it takes to minimize the time to aha moment. This is absolutely critical for any product design effort, but not many companies measure this if any at all when it comes to DX. I think Twilio and maybe Stripe are the only ones that may have had this as a key onboarding KPI.
Ultimately I think a majority of developers that are capable of implementing these things are quite technical and used to the general state of DX so they don't view bad DX as much of an issue unless its really terrible.
Lastly, I really wish error messages would just be super informative. For example getting something like this "undefined method `my_method_name' for nil:NilClass (NoMethodError)" still feels a bit cryptic to someone newer to programming, if you can also tell me the human readable variable that I used that caused this issue, the one that was nil, and the exact line (the stack trace purely by itself can be confusing) that little touch would go a long way. For example compare that error message to something highlighted in a different color that says "The variable you used called "contact" on line 87 was found to be nil, this is likely causing this issue". This way when you run into the error and are scanning the stack trace, the computer is telling you as quickly as possible what may be wrong, again for a novice since the way the original error is written for someone more experienced is likely succinct enough.
I couldn't agree more!
Why doesn't every error message have a link to a specific page with discussions, instructions, etc. Or maybe even a button you can click where the machine tries a best guess at an automatic fix?
> When was the last time you heard of a programming language discussed in terms of discoverability, succinctness, relevance, let alone beauty?
I mean, fairly often throughout the years. Particularly in communities for lisps, ruby, perl, python, C. More common 5-10 years back perhaps.
Relevance is a fairly common topic across the board. Discoverability is maybe the least common topic here, and one that's a pretty interesting one for PL design imo.
I haven't seen too many blog posts about these things lately, but they're frequent enough discussions in personal circles and in mailing lists/chats that this question seemed odd to me.
I disagree with the premise.
The problems the author identifies largely have already been solved. The solutions just haven't become universal, generally for reasons completely unrelated to the actual problem, and more to do with PR.
Interesting, what example would come to mind?
This line seems to sum up the core of the complaint:
> When was the last time you heard of a programming language discussed in terms of discoverability, succinctness, relevance, let alone beauty?
The ruby-lang mailing list used to be full of these sorts of discussions. If the author hasn't come across these factors being spoken about, that says more about the social value systems around the dominant ecosystems than about any fundamental complexity. Ruby lost the PR war to JS.
> Coding tools were around before UI/UX was a thing...
...yeah, no. At least, not in a way that makes the author's point. COBOL was an attempt to improve the developer experience. UI research predates JS by decades.
This paragraph approaches the complexities in JS as though they were a natural consequence of when JS was written, as though it wasn't possible to have done any better so we've all got to live with the best that was available then, rather than what we know now. That's just not true. JS even when invented wasn't a good language. Brendan Eich wanted to write a Scheme, and we'd all have been better off if he'd got away with it, and also if he'd had more than 10 days to implement it. The things we complain about in JS were commonly known to be bad at the time, they just didn't end up fixed for reasons entirely unrelated to the technology.
I wouldn't say it's fundamentally harder. It's fundamentally the same thing, because developers are users, and software development isn't the only domain where users have discerning tastes and needs that they want catered to. It's not the only domain where users will happily accept a steep learning curve for the years of use that will pay off that initial investment, or where shoving detailed information under the carpet can be a negative thing.
I think it would be arrogant to think of software development as exceptional in this sense, but it's certainly reflected in a lot of software designs. The reality IMO is that if you aren't working off a set of insights and observations like the one listed in the article—regardless of your users' domain—you aren't making enough of an effort to design for your users.
You think it is harder because you are not familiar with the depth of non-developer workflows. Developer hubris.
Indeed! Compare to CAD or Photoshop for deep, complex workflows.
I was wondering about this -- but how would CAD or Photoshop be more than a Context Free language? In particular, consider Feature Structure Grammars which are equivalent to CFGs and can really capture a lot of inter-context agreement without needing to be Context-Sensitive.
I certainly may be wrong though, all sorts of stuff ends up being Turing-Complete.
They can be, and afaict, are, typed context sensitive languages.
I wish we could scrap the term experience again and just go back to interface.
I connect experience with hyped concepts that are already forgotten today.
That said, tendency to decrease choice seems to only serve certain users. Other feel just as restricted as developers, which are also users, so the dichotomy should be questioned.
"DX" is indeed important. Myself and a number of other academics have been researching the topic for years.
See here for examples of publications in the area: http://web.eecs.utk.edu/~azh/publications.html
Also relevant, I wrote a blog post for students to get started in human factors in software engineering: http://web.eecs.utk.edu/~azh/blog/guidehciseresearch.html
Thanks for posting this! So happy to see some real academic interest in these topics!