And, this is a personal anecdote of course, but sometimes having fast machines can paper over problems that you should really be addressing: I had a coworker once who temporarily used a fairly (~7 years) old machine to do their development work until they could get something newer to work on. During that time I noticed that they would take a while to make minor fixes after reviews. I asked them why and they told me that they would make the change and then they'd build it and it was taking dozens of minutes–it turns out that there was a creeping dependency tracking issue in the build process that nobody had bothered to address yet because it didn't seem like a huge issue on our normal machines. But fixing it certainly got his turnaround time way up, and the CI machines really liked the change too ;)
Or you should -know- what specs your users are using (this is really not hard) and optimize for that.
You can get chrome to degrade network performance etc. It's not easy to downgrade machine performance tho.
I can count on one hand the number of companies I've worked at in the last decade who had in-house QA teams. Most companies just rely on the developers to test their (our) own code. The exceptions are banks, gamedev companies and big corp products (like Windows).
Most web software I've seen - especially in the startup world - just assumes that "it works well enough on the developers' specced out mac" means it works. The most tragic version of this I saw was our app looking awful on a ‘normal’ 19” desktop monitor. The problem was our designers did all their work from 27" imacs. They never thought to design and test on smaller displays. Or the designers placing adjacent colors that look great on color graded displays but are almost indistinguishable on the trashy 6 bit TN displays that were ubiquitous in offices a few years ago.
Even in companies with dedicated QA I haven't seen many tickets of the form of "the program runs slowly on this old computer we have". Sadly the creeping performance rot is mostly invisible.
Honestly, people running slow computers are probably just used to poor performance. I fully agree with you that we should do something about it but I don't see that happening unless it's a strong priority (feature X/y/z or performance? is usually the conversation).
Anecdotally: I'll never forget the first time I launched an app from an SSD and vowed I was never going back.
It's not as easy to do this as it could be, but it's not that hard either.
It doesn’t have to be either/or, it can be both. Developers absolutely should be using their own software frequently, seeing and feeling how well it performs; trying it out on a lower-spec machine is always a useful exercise.
That doesn’t mean you can’t also have dedicated QA. Developers shouldn’t have to be in charge of QA, but they should have some involvement.
Hypervisor / virtualization tech should mean you get very close to actual host OS performance. Otherwise this whole container thing wouldn't work so hot?
yes > /dev/null &
Repeat until slow ;)If you fail to make those consistently you end up with large infrastructure bills and unhappy customers eventually.
That is literally my job! You should have someone like that too, or else you'll end up shipping an Electron app like Discord does.
And get to dominate their market niche?
People got to realize while native apps are nice and all, there is absolutely zero evidence that going native vs cross-platform improves any metric a business actually cares about. Users are flocking to cross-platform apps in droves!
Vocal message board users screaming their preference for native apps are not in any way a significant share of the actual market for these applications.
I seriously doubt the use of Electron is what's driving that.
And it's also worth remembering that "cross-platform" ≠ "bundle an entire web browser in the app". Qt exists, as one of many examples of application frameworks/toolkits that are simultaneously cross-platform and native.
It is indirectly driving that.
Using Electron instead of other cross-platform frameworks like Qt allows one to work faster since you can apply webdev knowledge, and that directly leads to features that users actually care about being delivered faster.
Like, if you just need to spit out a prototype to hit the ground running fast, then fine, but long-term there are pretty clear benefits to using the right tool for the job.
The "work faster" argument is pretty dubious on its own anyway; yeah, maybe web devs ain't particularly productive with Qt or similar frameworks, but desktop devs absolutely can (and arguably should) be productive with such a framework.
There is no evidence to believe that users are concerned about RAM usage to any level that would move them over to competing apps.
> The "work faster" argument is pretty dubious on its own anyway; yeah, maybe web devs ain't particularly productive with Qt or similar frameworks, but desktop devs absolutely can (and arguably should) be productive with such a framework.
There are orders of magnitude more web devs than devs who are even just familiar with desktop GUI toolkits. Even if you can find people who are just as productive with (for example) Qt, they'd be (way) more expensive than the average web dev.
And, again, now you have less budget to spend on things that actually gets you users.
Do you know that writing code efficiently can also save money for the company? If you try to write code that burns bandwidth or CPU cycles without even attempting to optimize it, it's gonna cost you an eye at the end of the year.
Software optimized for the slowest hardware possible actually does saves money, which helps also to reach the goal of doing more money, since more hardware is being supported, not only the cutting-edge one.
I've seeing now this infamous trend also on Android Phones boosting 90-120 Hz screens. If applications get optimized to run on these phones, whose CPU is also strong, obviously older phones are suffering and become sluggish.
Seriously, try to write code as your target platform is a Raspberry Pi 1 model. If works on it fast, it's going to be fast everywhere
Your job isn't to build the smoothest, most optimized, experience for your customers. It's to build features customers want or need, which in turn create value for your employer. This generated value pays your paycheck. Of course if you do a bad enough job, it's counterproductive, but to the extent it's not bad enough your job is to focus on features and development and not on optimization. Once performance is bad enough, your job becomes optimization -- via instrumentation, measurement and profiling.
Every minute you spend waiting on a build or waiting for an app to launch or waiting for some long-running task to finish is a minute you're not creating value. Compounding this problem, it impacts morale if you spend your day looking at spinning beach balls while your boss won't get off your back about when you'll be done.
This is your employer's perspective, more than likely.
[edit] Another way of looking at this is to say that you and your customers are doing different jobs. You have all the extra cpu and memory overhead of developing a product — all they have to do is run it. By definition your workload is more demanding.