For instance, searching for "Java read a file", I find a bunch of examples. And they mostly are good enough, call out when they aren't, and provide useful links.
Meanwhile, navigating Oracle's website is a nightmare. https://docs.oracle.com/javase/tutorial/essential/io/file.ht... exists, and is actually pretty helpful, but the only way I was able to find it was by Googling "oracle java read a file". Searching Oracle's own doc site (itself hard to find for a given version of Java) with the same string (i.e., 'read a file') turned up nothing useful ( https://docs.oracle.com/search/?q=read+a+file&category=java&... )
I do a lot of GCP work anymore, and the amount of devs who make service authentication way more complicated than it needs to be is really freaking high. It's clear that the people who kicked off the project are pulling blocks from SO rather than RTFM because the FM has a simple decision tree for how to use authentication. Reading through the docs could have saved these projects each hours of time, simplified deployment and reduced the security risks associated with passing around json auth tokens (or worse: checking them into VCS.
- Beginners do not understand doc's jargon, or they need to learn too much stuffs to understand it that they'd rather quit. So they end up relying on third party sources/implementations
- No English knowledge and no translations
I have a spanish-speaking group where me and others teach beginners programming, and many of them cannot use the docs, whether because of a language barrier, or a lack of foundational knowledge, and all they care is to get things done, otherwise they'd get discouraged and abandon it before even starting
People generally just want to cut corners and 'be done with it' which simply results in lower quality.
Everyone of your users will start off as a beginner, it behooves project creators to ensure that beginner documentation is great.
“But the docs are poorly written” you say
“But the docs aren’t maintained well” you say
Read the docs. Sometimes it isn’t supposed to be easy. Or fun. Or fit together well. But that’s the job.
RTFM is the best technical advice I’ve been given and can give.
In practice, difference between it and Oracle docs was negligible. Hell. Their auto-gen code was more self explanatory than Oracle Java Docs.
new Doer(UnrelatedFactory.create("foo", new WTFBar(), 5, null, null), RandomEnumFromADifferentPackage.MEANINGLESS_VALUE, true).do()
But I'm not really convinced such documents are helpful. I mean, technically they're better than nothing (as in, literally better than nothing), but the difference between nothing and those docs is so small it approaches nothing ('as the worthlessness of the documentation approaches infinity the value of reading them approaches nothing', as it were)
It's a bit more complicated than that. I do read the docs, but using the web is essential for me, for a few reasons:
- nowadays, a dev doesn't work on an area that can be included in a single book; one touches several different frameworks, and reading that docs for all of them can be just unfeasible
- the docs may not be good; you may be surprised that even the Linux API doc can be really bad
- there may be issues that require workarounds that are not described in the docs
- best practices may not be obvious even when reading the docs
- one can't assume that in a big doc, it's obvious where the solution for a problem is, which may force to read way more than expected
I think a dev can "google more than RTFM'ing" and still produce a good solution, if they intend to - that is, to process the information rather than blindly copy/pasting.
Thinking about it, the major problem seem to me unprocessed copy/paste, rather than "googling instead of RTFM'ing".
> nowadays, a dev doesn't work on an area that can be included in a single book
Which is terrible because it means that there is less specialization and its benefits for productivity.
> the docs may not be good
You get what you pay for.
> there may be issues that require workarounds that are not described in the docs
You get what you pay for.
> best practices may not be obvious even when reading the docs
You get what you pay for.
Don't get me wrong, OSS is amazing, but in commercial software, you also pay for support & documentation. But since nobody wants to pay for the software anymore, we end up with using Google instead of reading the documentation. And this becomes self-fulfilling prophecy, why provide good documentation if nobody is gonna read it?
It would be ridiculous if somebody wrote an article stating that "googling it" is the most important skill for a mechanical engineer or a doctor. But for software engineering, it is acceptable. Looks like software engineering is becoming less and less about engineering every day.
Addendum: SW developers also seem to write less, comments or other human-readable explanations. This might also contribute to less appreciation of good SW documentation, and contributes to a worrying trend of being less literate.
I am right now working on a project because of the googling stuff has led to the worst practices at the moment. I have worked on many projects where people haven't bothered reading the docs and I have seen the same pattern for over a decade.
Whereas if they had read the docs the best practices would have been immediately apparent. So that argument hold no water with me.
> one can't assume that in a big doc, it's obvious where the solution for a problem is, which may force to read way more than expected
I want people reading more. Which was part of my overall point. People don't read enough. They go for the quick solution (which is copy and paste from stack overflow) produce something which is usually quite bad and then you normally have to redo that work because the moment you try to change something (due to changing requirements) the whole thing falls to pieces.
I've been guilty of it too. It is a bad habit that is produced through laziness.
I literally wrote docs at a previous place and people kept on asking for help on IM. I wrote the document so I wouldn't have to answer the same questions many times. People getting into bad habits and not looking for or writing docs literally causes a hell of a lot more issues that you would see at first glance.
> Thinking about it, the major problem seem to me unprocessed copy/paste, rather than "googling instead of RTFM'ing".
I would rather have a unprocessed copy and paste then what you get in reality which is a "copy pasta". There are minor bits changed which make it difficult to recognise where it may have originally come from, which makes it difficult to know the original developer's intent.
For example. If you are using the babel compiler and its plugins you will fine lots of answers that are wrong on Stack overflow because they were written when Babel 6 was the major version was popular. Whereas reading the docs (the doc site is quite good IMO) will get you the correct answer every time but comes much lower in search results. Using the babel 6 suggestions can lead you to into some sort of NPM hell.
Yes, a large portion of developers just copy/paste from SO. However, many people also use the results of what they find on Google to learn concepts that the actual docs for just aren't that good.
Sometimes when you need to get something done and you find something that works as is, it is easier to test it and move on to something more important.
We don't have infinite time.
But on the second point, I find myself having to use the g! Operator of ddg to get the result I need.
Recently it was looking for a workaround I knew existed for a documented defect. ( weirdly not documented on the bug tracker of the library ... go figure )
Google gave me the result, Duck Duck go was serving me the doc or the bug tracker.
Hmm. Actually maybe Google just remembered that from my previous search.
I cannot offer that to the users of my own docs, because I know how outdated it was...
I really wish people would just read the docs instead of just googling. I am working on a large typescript solution and almost every error in the application is due to the original developers not reading the docs and instead just googling stuff from stack overflow.
Also I am against the promotion of google generally. The search engine isn't that much better than Duck Duck Go anymore (especially for dev stuff). Also we should all know the problems with Google's defacto monopoly on search.