Why I Skip Stack Overflow

StackOverflow has its flaws but compared to the competition it's heaven. Moreover, when reading OP's 'Computer systems are designed to be figured out from their manuals or APIs' I stop reading. Sometimes systems are so new there isn't a manual yet (eg when Node started) or you just deal with complex matters where multiple systems are interweaved.

I learned tons of stuff on SO I would have never figured out myself. Also, how comments and answers are conceptually and visually separated is just very smart and the pain point of every other Q&A site.

I go straight to StackOverflow with no shame. I rarely use the code snippets as-is, but even pointing me to a useful library that I didn't know about can be helpful.

I appreciate the exercise of learning how to use a system from its manuals. I practically memorized the manuals for every successive version of Turbo Pascal up to around 1990. I know every chip in the 1982 National Semiconductor Linear Databook.

But Turbo Pascal was also completely self contained, and came from a single source. Today, I don't even know what the "manual" for something like Python and its ecosystem of add-ons, really means. But whatever it is, a manual needs an index, and for me that index is Google/StackOverflow.

When I run into an issue, and happen to land on a stack overflow page when searching on it, I don't just try and find a solution that lets me do what I want to do. I try and figure out -why I'm having difficulty- in the first place.

This might be because I don't know the language/library/whatever, or have forgotten some key bugaboo in the stack I'm using, or that its documentation hasn't been updated, or that the holistic knowledge I need to get a useful example isn't really spelled out anywhere canonical (all that exists are descriptions of API endpoints, not a manual or tutorial that tells me how to bring it all together). In which case, a simple "here's how you do it" is -entirely sufficient-; SO can deliver this.

But it also might be because I have a fundamental misunderstanding, as per the OP's example. In which case I read the answers on SO, or anywhere else, -to correct that understanding-. And frequently, SO delivers on that. While there may be a one liner "Just do XYZ" to hack around it, there is almost invariably (and in fact, it's usually the highest rated) an answer from someone explaining the code or idea in question, and the right way to do it. Again, SO can deliver this.

This decision of which sort of answer you're looking for isn't due to a specific source of answers, but what sort of answer is good enough for you. If you have an issue of the latter variety, where you have a misunderstanding, but the former kind of answer is sufficient for someone, where they just are enabled in that misunderstanding, rather than correcting it, then that speaks ill of the person accepting that answer, not of SO enabling that (along with the correct) kind of answer.

The second issue is with understanding. I hold a firm rule, never commit code that you do not fully understand. Because when it breaks, you will have no idea how to fix it. Too often I see developers copy/paste directly, and just change variable names. You aren't writing code, you're playing dev Lego.

My final issue is their draconian insistence on what they consider subjective question locking. When I'm making a technological choice, I'm looking for real world usage, and comparisons between frameworks. These questions can be asked and answered in non-subjective ways; yet even when done so, anything with the word "compare" gets locked. This would be the real value to me, as my research time is very limited, and being able to exclude frameworks and tools based on objective comparisons of feature sets or capabilities and limitations would be highly valuable to me.

Sometimes you get stuck on things that are blockers and time consuming but there isn't a lot of learning value there.

Saving time on these kind of blockers means i can spend more time on learning concepts and higher level things that help me become a better coder.

It will be interesting to see how this plays out, but I suspect that like the Luddites and their looms, augmented human intelligence via computers will win out in the end. People will simply find it more efficient to tackle learning from a agile perspective using search and crowd sourcing to grab what you need when you need it rather than doing a lot more reading up front and later sifting through what they need and don't need (YAGNI for knowledge?).

The example given by the author about moving buttons in iOS tries to make the case that there are unknown implications to finding a "hack" to solve a tough problem. I look for this type of information on SO all the time – usually not to implement the exact solution but to learn how others have solved similar problems that will help me think about the specific issue I am tackling. That's how I learn. Not for everybody, and I recognize we all do things differently.

We were working on a legacy C# project where the initial architecture was such that a RAKEFILE (ruby-make) was used to make a call to CSC.exe (the C# compiler).

"It's easy, we use Ruby to compile our C# so we can use wildcards .cs" (to say nothing of the problems caused by the loading of '.DLL' from the bin folder on app start)

Over time, when I was finally comfortable with the project, and after days of trial and error, I added CSPROJ files for each of the projects and was finally able to compile, run, and debug in visual studio.

When I tried to commit the CSPROJ files into git, he deleted them on the next commit.

When I asked why he removed the files is was a gongshow. "Why in hell should I allow you to put your shit in my repo?" He snapped.

When I explained that using intellisense & debugging were both incredibly valuable.

He countered with "why don't you just memorize the namespace (because that is what better programmers do)"

And "why don't you just use print statements to inspect values? (because that is what better programmers do)"

Skipping stack overflow is a similar sort of self hamstringing that I can't get behind.

Getting the job done quickly is part of what we're paid for. (Caveat* optimally quickly of course, there are plenty of times when taking a bit longer pays off in the long run)

Anyways, TL;DR skipping stack overflow is dumb.

Yesterday as I was fooling around with Writing GNU Emacs Extensions. One of the great things about that book is that Glickstein figures [correctly in my case] that someone learning to write extensions might not have a good handle on other aspects of Emacs...such as help. It hit me how googling tends to drop me into the emacs manual or other high level references and how impossible it is to drill down on specifics in the way the Emacs help system allows. I mean I can google up a list of the top 150 Emacs keyboard shortcut, but I can't google up "C-x n".

While I love StackOverflow, the overhead of "What does 'C-x n' do?" is astronomical...even if such questions were appropriate.

I've trained myself to break flow and look at ads every few minutes because Google seems great. I'll even google knowing it will have some StackOverflow links under the ads at the top instead of going to the site directly. What I won't get is the great local resource that sits on my machine.

I guess basically boils down to conversation vs searching. In the last couple of years (maybe more) I have seen less searching and more asking questions in general. I think a large group of mature technologist know how to use Google and prefer it over human intervention... ie a conversation. That is the initial tech culture was RTFM. But there is a growing population particularly where English is not their first language where they prefer to ask over search. I believe this is also culture thing as well. Perhaps a more social culture.

In the US we are so used to Google but in other countries social media, messaging apps, and improved connectivity are really pushing the "conversation interface" and appears to be preferred heavily over the "search interface". That is people seem to want to ask and not search.

I honestly don't know for sure which "interface" is more efficient or better but clearly given Siri and various other AI conversation thing coming out the conversation approach is popular.

Lol. If only every API was designed and documented well.

Have you tried to code angular2 or something else new? Documentation simply does not exist.

StackOverflow is the de facto documentation source for a lot of things.

On lots of occasions it is not a shortcut, it's simply the only place you can get an answer.

If SO can help me solve some dumb CRUD problem more quickly or navigate a rather poorly written API (android, iOS, azure, and, well, most APIs actually), then I'll take that route. It never ceases to baffle me how many developers think esoteric language or library knowledge makes them good at their job. You're missing the forest for the trees guys and gals!

As another commenter said if I find that the SO answer provides a counter intuitive solition or gives me some hint that I'm thinking about something in the wrong way then I will of course chase down the "why" of what I'm doing.

Funny thing is there are a lot of sites that just republish the SO post. Also, there's the reverse where you find a really good blog post in an SO answer.

>so you copy and paste code fragments from up and down the page, in a vain attempt to shut the problem down

I almost never do this. I read the snippets to get the idea of why they would work which helps me understand the problem better.

>Learning to read an API or manual will save you time.

It's even better when the library/framework you're using is open source so you can just look at what the code is doing vs. what they say it's doing.

Source?

A lot of software nowadays implicitely acknowledges the existence of SO or even open recommends to use this source.

I get that a lot of the time you get your answer, but reading the answer and understanding it is why Stack Overflow is so beneficial. Yes it does save time but it should be used to learn.

Last time I saw a documentation I would consider beautiful, comprehensive and really well written was back in the OpenVMS days. You can see something distantly resembling that glorious style if you type 'help' in gnuplot.