We hadn't got that far with our use of RAML while I was there and our primary concern was providing external documentation to possible end users of the API. It was all going to be manually generated and maintained and I was glad when they hired a real product manager to deal with it all.

Yeah, our previous documentation was hand generated but either the dev that did it hid this from people, or they just didn't understand what he was doing, or what. He was "using swagger" to generate his docs from django rest framework endpoints, but basically what this meant was... you made specially formatted comments and the docs were made out of those.

RAML is not as good as it should be. We're using 0.8 because almost all the tooling that exists supports that. 1.0 is better but imo still not that well thought out. Everything else I looked at is just as bad or worse though. It's a hard problem.

Without really meaning to, I wrote a framework that we're using. It just sort of happened. It's completely self documenting when done right (so about 95% of our endpoints are self documenting, the rest are not).

I hand wrote some bastardized swagger for documenting our API on 3Scale and it was a pain in the ass. But the more I looked at automated documentation solutions based on tags in the code or what not I never found any that weren't the kind documentation written for people who already knew what an API did. It's a lot of work for someone and we never found anyone to own it. Probably because we never really opened our API to customers, when we did I would write up documentation and examples of the calls and we'd use that.

It is pretty clear the inverse is true. What is really interesting is how relatively large userbases can sustained by basic off the shelf tools. (Most B2B SaaS products for instance.)

Its clearly the last 5% that takes 95% of the effort and makes these businesses viable. Not slapping together the first incredibly buggy web spider in Python.

Whenever I visit bloomberg https://www.bloomberg.com/
my computer fan starts going crazy. Is there a plugin or something you know about that will help me figure out how to stop whatever is causing this ?

I have uBlock installed for ads already.

wow. i tested this (and i have ublock, ublock origin, and privacy badger installed), and CPU of my macbook shot to 300%, and then it went to sleep, presumably because it was overheating?

that is just unreal. how is a major news outlet that incompetent?

^ They've been aware of that for a long time, the site even displays a warning message for me that blockers may cause performance issues on the site. I've never experienced any though, I always thought that was just to get people to disable their blockers.

Wanted to loop back to this because I think it's interesting.

Why do you suppose this is happening? Is there an influx of college grads or H1Bs taking all the jobs? What about the "talent shortage" is different now that it can't be exploited?

This post got me thinking a bit though. I fee like, if you can afford to survive in SF for one year unemployed, and you can afford to pay for that course, that sort of allows for 2 options.

The first option is as you stated, seek employment for a year or more, but the second option is to consider using the environment and time to build up a business. IME, the work taken to apply to as many jobs as you can in SF (which can be about 200 / week) -vs- just going on your own is probably about equal.

Mind that when I say "equal," I mean small enough that you can build out a product alone without the need to get $5M and hire on 5 employees to launch. It's a tough balance to find something small like that, but it can be done.

It's odd that chrome itself doesn't protect you. In the same way a page triggering multiple popups prompts a checkbox allowing you to kill all further popups, a page consuming too much CPU should be automatically paused, requiring your permission to continue.

Isn't Chrome sand-boxed? What safe mechanism would they use to read CPU usage?

This makes me think of the old YouTube article where they sped up YouTube and found that the average load times actually ended up slower. This was because the page speeds were fast enough for the people with slow computers and slow connections to use without a browser / computer crash or infinite wheel spin. At that time, YouTube was always fast enough for residents of big city, but not fast enough for middle-of-nowhere with slow connections.

Is it really be possible to figure out what is considered "too much" CPU from computer and user to user? People, me included, are still using computers with old-gen processors (i5) or cheap single-core computers bought from Best Buy, which in stark contrast to the coffee-shop Apple user. We all use computers differently as well. I run a very lean system, but then you have some people running 12 dockers, Slack, and Spotify, plus a dev environment. Others will just have YouTube and MS Word running.

Bloomberg loads fine on Linux, both FireFox and Chrome. I have no plug-ins on Chrome.

There are a lot of mature boot camps churning out people right now which I'm sure it's a factor.

But I think it's just that 3 months was never enough time to go from zero to employable.

Even 3 years ago when I graduated there was a pretty stark difference between the top third of the class and everyone else (and even clear differences within that third). The common elements of the top third were that we had all dabbled in programming naturally and we're information sponges. The type that reads HN, reads docs for the latest js libs, takes moocs for fun etc. I think those types still have little issue getting jobs.

Maybe those types were early adopters and the average boot camper is a little weaker today as a result.

We just hired two people out of boot camp. One from Dev bootcamp and one from AA. The former was a referral and has a great work ethic but just did ok on our interview. The latter worked in a product management capacity with very limited code exposure to whet his appetite and a year ago started doing hackathons, he crushed our interview.

AA guy is fantastic and nearly fully autonomous after just a few weeks. The DB guy works really hard but still needs quite a bit of hand holding.

i don't know the answer to this



yes, it is. these are websites. i don't want them ever turning on my fans without my permission. it's really not a hard problem, even with vastly different specs. plus it could be configurable. it could be opt in only, etc. but it's crazy that there isn't a way to do it, unless it's technically not possible for some reason.

Software engineering is the only engineering field where reusable artifacts that are useful far beyond the context in which they were invented are routinely created as part of one-off projects. Consider how many products and open-source projects came out of Google, Microsoft, Facebook, Twitter, LinkedIn, etc, not as indepedent projects, but simply as necessary elements for their other existing engineering efforts. What drives this more than anything is the self-similarity across the whole production cycle and supply chain. Not only do we make software, but the components we use to build software out of are also software, the broader product our components are part of is also software, the tools we use to build software are software, the tools we use to communicate are software. This allows for coordination, control and collaboration across the whole supply chain to a degree unheard of in any other field.

Because they don't work on easy problems over and over again on extremely large budgets on a fixed scope for a very long period of time. Failure here is a meaningless concept - from a microeconomic perspective, the prices of input factors are generally going to be set such that projects are on average EV-neutral, which means any industry where outsized returns are possible will have a higher failure rate. So all that really means is that the median successful project in software engineering has a higher rate of return.

That's a total non-sequitur but:

http://science.howstuffworks.com/sci...cal-outlet.htm

It's pretty incredible that this is a problem in 2017.

"Using too much cpu" is a weird thing. That's something for your OS to enforce, not by each application themselves. A quick google suggests a homebrew of cputhrottle for Mac will do the job. renice is a different option if you don't have a poorly designed / gunked up with dust cooling system and rather want to allow a process to use as much cpu as it can without making other things unresponsive.

A computer that shuts down from overheating at 100% CPU usage is a faulty computer.

I'd guess most people would want as much CPU as possible for the immediate task at hand, generally.

Or you could just underclock until it's running at a sensible temperature in bios.

People, not processes. And people are expensive because they are valuable - are you saying software engineering is not engineering because it generates too much economic value? Either way, your argument has devolved into finding less flattering comparisons depending the context. When you're talking about how "engineering" is also impressive and complex and whatever, you're talking about the occasional cutting edge stuff, which as a group was also super-expensive, unpredictable and had high failure rates. When you're talking about how "software engineering" fails all the time, you're comparing it to routine projects that have virtually no real engineering involved except reusing existing design and someone signing off. One caveat here is that because a very high degree of customizability and reusability is often built into good software products and a seemingly infinite number of features that all serve a very small percentage of users can be crammed in without compromising usability, any new software requires a higher bar for differentiation than physical products - this means every new software project is much more of an R&D project than non-software projects, where small differences can be enough of a basis for a product. Or in civil engineering, no real difference is even necessary, since even a copy would be a new project.

And there's no such thing as "the industry" - the broad scope of software engineering is such that the most valuable retailer in the world is a software company (Amazon), the most valuable advertising company in the world is a software company (Google), the most valuable electronics company in the world is a software company (Apple) and so on. Software engineering is so profitable that practically every major company is trying to do more of it and the world as a whole is trying its hardest to satisfy this demand.

I recall fan-spinning and even computer freezing happening a lot 2 years ago.

Nice thing about larger JS / CSS frameworks is that they are much faster and standard. I guess other practices have become common, but the obnoxiously slow or stalled out website on a decent connection isn't nearly the problem it used to be for me.

And a lot of websites seem to load just fine on my phone. I know that is partly because Opera mobile doesn't download every resource, but responsive frameworks with baked-in media rules are making a large impact as well, I reckon.

it happens fairly often for me. eg, spanishdict was crashing safari on my iphone 6. i installed an ad blocker and now it works fine.

Congrats!!!

I think at this level, "communication" is probably a better all-encompassing term than "UI" right? UI is just how your product communicates with the user.

Now that I'm at a bigger company, I also find the communication structure within an organization quite interesting. Most importantly, how people decide who to talk to about certain things and also how leaders decide who should know what and also what they do to ensure that. It seems that beyond talent, these kinds of factors largely determine the productivity of software engineering teams at large.

there's overlap with the concept of "communication." but it's subtly different in my mind -- some specific subset of "communication." the skill is in anticipating how some ambiguous thing without clearly agreed upon social meaning will likely be interpreted or misinterpreted, and correcting in advance for that. "UI" (in the screen sense) is a good metaphor because the ambiguity is turned up so high. Ambiguity exists in language, too, of course, but we also have dictionaries, years of schooling, practice since infancy, and so on. even with our screen addictions, and even acknowledging some fringe academic attempts to treat it as language, i think UI is fledgling by comparison.

If you want an interesting take, check out "Making Things Happen" by Scott Berkus. The whole book is fascinating, but the chapter on politics is worth the entry fee.

FWIW, I've only been in companies up to 100 employees.

sounds like conways law in action: https://en.wikipedia.org/wiki/Conway%27s_law

i think this law, and your observation, are deeply true

UX is the term you guys are discussing.

Think it's Chrome who have an interesting feature proposition for future where you can assign a max CPU usage to a frame on your site. IIRC it's something the web dev sets. Think this will probably be used as part of an advertising platform feature to improve web ads, can't see much else use for it.