13 Things People Hate about Your Open Source Docs

open source anger

Most open source developers like to think about the quality of the software they build, but the quality of the documentation is often forgotten. Nobody talks about how great a project’s docs are, and yet documentation has a direct impact on your project’s success. Without good documentation, people either do not use your project, or they do not enjoy using it. Happy users are the ones who spread the news about your project – which they do only after they understand how it works, which they learn from the software’s documentation.

Yet, too many open source projects have disappointing documentation. And it can be disappointing in several ways.

The examples I give below are hardly authoritative, and I don’t mean to pick on any particular project. They’re only those that I’ve used recently, and not meant to be exemplars of awfulness. Every project has committed at least a few of these sins. See how many your favorite software is guilty of (whether you are user or developer), and how many you personally can help fix.

1. Lacking a good README or introduction

The README is the first impression that potential users get about your project. If the project is on GitHub, the README is automatically displayed on the project home page. If you get it wrong, they might never come back. You want to grab the reader’s attention and encourage them to continue investigating the project.

The README should explain, at least:

  • What the project does
  • Whom it’s for
  • On what hardware or other platform it runs
  • Any major dependencies, such as “Requires Python 2.6 and libxml”
  • How to install it, or a pointer to in-depth directions

All this has to be written for someone who never heard of your project before, and may never even have considered something like your project. If you’ve got a module that calculates Levenshtein distance, don’t assume that everyone reading your README knows what that is. Explain that the Levenshtein distance is used to compare differences between strings, and link to more detailed explanation for someone who wants to explore its uses.

Don’t describe your project in relation to another project, such as “NumberDoodle is like BongoCalc, but better!” That’s no help to someone who’s never heard of BongoCalc.

2. Docs not available online

Although I haven’t seen any studies on the topic, I’d bet that 90% of documentation lookups are done with Google and a browser over the Internet. Your project’s docs have to be online and available. Given that, it’s embarrassing that my own project, ack, would neglect having the docs available where most people would look for them. My assumption was based on my own use case, that if I want to know how a command line tool works, I’ll check its man page.

How was this brought to my attention? Users wrote to me asking questions that were answered in the FAQ, which made me annoyed that they weren’t reading my FAQ. Turns out they were looking at the website, but I hadn’t posted the FAQ there. It’s an easy mistake to make. I’m close to the project and I never actually use the FAQ myself, so I don’t notice it missing online. Many problems fall into this trap: Authors not putting themselves in the users’ shoes.

3. Docs only available online

The flipside of this problem is to have the documentation only online. Some projects do not ship the documentation with the project’s deliverables, or include a substandard version of the docs.

The search engine Solr, for example, has an excellent wiki that is effectively the project’s documentation. Unfortunately, the docs that ship with the download are 2,200 pages of autogenerated API Javadocs. The only documentation for the end user is a single page tutorial.

The PHP language doesn’t ship with any documentation. If you want the docs, you have to go to a separate page to get them. Worse, only the core docs are available for download, without the helpful annotations from users (see “Not accepting user input” below), and they’re not in the same easy-to-navigate format that’s available online.

Open source projects can’t assume that users have Internet access when they need docs. Airplane mode still exists. Even then, you don’t want the user to rely on your project’s website being up. At least twice over the past few months I’ve found the Solr wiki to be down in the middle of the workday while I was hunting for information on a tricky configuration problem.

One project that gets it right is Perl and its CPAN module repository. Documentation for each module is available at either search.cpan.org or metacpan.org in an easy-to-read hyperlinked format. For offline needs, the documentation for each module is embedded in the code itself, and when the module is installed on a user’s system, local documentation is created as man pages. Users can also use `perldoc Module::Name` to get the docs from the shell. Online or offline: It’s your choice.

4. Docs not installed with the package

This problem is usually a failing of the package creators, not the project authors. For example, in Ubuntu Linux, the documentation for the Perl language is a separate, optional package from the language itself. The user must know she has to explicitly install the documentation as well as the core language or she won’t have the documentation when she needs it. This trade-off of a few megabytes of disk space at the expense of documentation-at-hand on the user’s system serves everyone poorly.

5. Lack of screenshots

describe the imageThere’s no better way to grab the potential user’s attention, or to illustrate proper usage, than with judicious screenshots. A picture is worth a thousand words. That’s even more important on the Internet because you may not get the reader to read more than a few hundred of your words at all.

Screenshots are also invaluable for the user following along with the prose, trying to make something work right. A screenshot lets him visually compare his results to those in the docs to reassure himself that he’s done a task correctly or to easily find what’s not right.

It’s becoming more common to have videos on the website giving an overview of the project, and those are great. So are in-depth videos that show the steps of a complex process. The Plone project, for example, has an entire site dedicated to video tutorials. However, videos can’t take the place of screenshots. A user wants to see quickly what the screens look like without sitting through a video. Videos also don’t show up in a Google image search, as screenshots do.

6. Lack of realistic examples

For code-based projects, the analog of screenshots are good, solid examples of the code in use. These examples should not be abstract, but direct from the real world. Don’t create throwaway examples full of “demo name here” and lorum ipsum. Take the time to create a relevant example with a user story that represents how the software solves a problem.

There’s a reason we have story problems in math class: They help us apply what was taught.

Say I’ve written a web robot module, and I’m explaining the follow_link method. I might show the method definition like this:

   $mech->follow_link( text_regex => $regex_object, n => $link_index );

But look how obvious it becomes when adding some reality in an example.

# Follow the 2nd link matching the string “download”

   $mech->follow_link( text_regex => qr/download/, n => 2 );

The abstract placeholder variable names $regex_object and $link_index now have some grounding in the mind of the reader.

Of course, your examples shouldn’t just be brief two-line snippets. As Rich Bowen of the Apache project puts it, “One correct, functional, tested, commented example trumps a page of prose, every time.”

Show as much as you can. Space is cheap. Make a dedicated section of documentation for examples, or even a cookbook. Ask users to send in working code, and publish their best examples.

7. Inadequate links and references

You have hyperlinks. Use them.

Don’t assume that because something is explained in one part of the docs that the reader has already read that part, or even knows where it is. Don’t just mention that this part of the code manipulates frobbitz objects. Explain briefly on the first use of the term what a frobbitz object is, or link to the section of the manual that explains what a frobbitz is. Best of all, do both!

8. Forgetting the new user

It’s sometimes easy when writing the docs to write them from the perspective of you, the author of the software. New users need introductory documentation to ease them in.

The introduction should be a separate page of documentation, ideally with examples that let the new user get some success with the software. Think about the excitement you feel when you start playing with a new piece of software and you get it to do something cool. Make that happen for your new users, too.

For example, a graphing package might present a series of screenshots that show how to add data to a file, invoke the grapher, and then show the resulting graphs. A code library might show some examples of calling the library, and then show the resulting output. Keep it simple. Give an easy win. The text should introduce terms at the appropriate places, with links to more detailed documentation about the term.

A separate document for these sorts of introductory ideas gives the user a quick understanding of the software. It also keeps the introductory explanations out of the main part of your docs.Not listening to users

9. Not listening to the users

Project owners must listen to the users of the documentation. The obvious element is listening to suggestions and requests from people who are actively using your software. When a user takes the time to mail or post something like, “It would have helped me install the program if there had been an explanation or links to how to install database drivers,” take that message seriously. For every one user who emails you about a problem, you can expect that ten others don’t say anything but still have the same problem.

Just as important, however, is listening to user problems, and considering the reasons behind them. If people frequently have trouble figuring out how to perform bulk database updates, the first course of action is to add a question to the FAQ (you do have an FAQ, right?) that addresses bulk database updates. However, the question may also be an indication that the section on database updates isn’t written clearly enough. Or perhaps there isn’t a pointer to that section from the introductory overview document, so your users never know to read the rest of the manual.

Besides helping more people discover how useful your project is, this also eases frustration on the part of the project’s existing community. If your mailing list, forum or IRC channel is filled with people who ask the same “dumb” (or not-so-dumb) questions all the time that everyone gets tired of responding to, recognize that these are the Frequently Asked Questions, and putting the answers in a single find-able spot helps everyone focus on the fun stuff.

Keep an eye on user questions in outside forums, too.  Search sites like StackOverflow regularly, and set up a Google Alert for your project name to be kept aware of how your project is being discussed on the Internet.

10. Not accepting user input

If your project has a large enough user base, it may make sense to incorporate user comments directly into the documentation. The best example I’ve seen of this is for the PHP language. Each page of the documentation allows authenticated users to add comments to the page, to help clarify points, or to add examples that aren’t in the core docs. The PHP team also gives the reader the choice of displaying documentation with or without user comments.

As useful as this is, it requires maintenance. Comments should be weeded over time to prevent overgrowth. For example, the PHP documentation page for how to invoke PHP from the command line includes 43 comments from users dating back to 2001. The comments dwarf the core documentation. The comments should be archived or eliminated, with the most important points incorporated into the core documentation.

A wiki is also a good approach. However, if your wiki doesn’t allow the user to download all the pages in one big batch (see item #3 above), then your users are at the mercy of your Internet connection and the web server that hosts the project.

11. No way to see what the software does without installing it

At the minimum, every software project needs a list of features and a page of screenshots to let the curious potential user why she should try it out. For the user shopping around for software packages to use, make it easy for him to see why it’s worth the time to download and install it.

Pictures are a great way to do this. Your project should have a “Screenshots” page that shows real examples of the tool in use (see item #5 above). If the project is purely code, like a library, then there should be an example page that shows code using the project.

12. Relying on technology to do your writing

Too often, software authors use automated documentation systems to do their work. They forget about the part where they have to write prose. The automated system can make things easier to maintain, but it doesn’t obviate the need for human writing.

The worst case of this is changelogs that are nothing but a dump of commit messages from the version control system, but with no top-level summary that explains it. A changelog should list new features, bugs fixed, and potential incompatibilities, and its target audience is the end user. A commit log is for people working on the project itself; it’s easy to generate, but it’s not what users need.

Take a look at this page from the docs for Solarium, a PHP interface to the Solr search engine. First, the disclaimer takes up the top half of the screen, giving no information to the reader at all. Second, literally nothing on the page is any more descriptive than a list of the function names. The “METHOD_FC” enum means “Facet method fc”. But what is FC? There is no explanation of the different facet methods, nor links to how one would find out. The automatically generated pages look nice, and they may feel like documentation, but they’re really not.

13. Arrogance and hostility toward the user

The attitude of RTFM (Read the Freaking Manual) is toxic to your project and to your documentation.

It is the height of arrogance to assume that all problems that relate to someone not knowing how to use your project are the fault of the user. It is even worse to assign a moral failing of “They’re just too lazy to read,” which you can’t possibly know. Most of us have encountered these attitudes as users. Why would you inflict them upon the people who want to use the software you create?

Even if it’s provably true that users could find their answers in your documentation but they aren’t doing so, it’s foolish to assume that it’s because the user has failed in some way. Maybe your documentation is poorly written, or hard to read, or presented poorly on the screen. Perhaps you need to improve the Getting Started section (item #8 above) that explains what the software aims to do Maybe some bits of information need to be repeated in multiple parts of the docs. Maybe it’s not made clear to the reader where to find certain knowledge tidbits.

You know that new users of your software come to your project knowing nothing. Your project documentation team can do its best to ensure that ignorance is easily curable.


I’m sure you have come across many of these problems with docs, and I hope there are some you haven’t thought of. Let us know about the problems that bother you in the comments below. I don’t mean to point fingers at any given project’s examples, since every open source project has some problems or another.

Most of all, I hope that if you recognize a problem in your documentation on projects you’re involved with, that you consider this a nudge to take action to improve the situation. Fortunately, improving documentation is an ideal way to get newcomers involved with your project. I’m often asked “How can I get started in open source”, and I recommend documentation improvements as a good way to start.

Make it as simple as possible for contributors, whether novice or veteran, to know where your docs need help. Create a list of tasks, possibly in your bug tracking system, that explain what needs help. Be specific in what your needs are. Don’t just say, “We need more examples.” Create specific tasks, like “Add example code of how to do Task X,”  “Add screen shots of the report generator screens,” or “Add dependency information to the README.” Contributors want to help, but are often stymied by not knowing where to begin.

Docs aren’t the most glamorous part of any open source project, and for most of us they’re not fun, but without good documentation, users aren’t served as well as they could be, and your project will suffer in the long run.

See also:



  1. i disagree with point 4. documentation should be in s seperate package, especially if it is big. perl is a good example, often installed in embedded systems where there is really no space for docs. 
    the packaging system can make the user aware of a docs package via suggestions or recommendations (in case of deb packages) which i’d consider good enough to find them. better than forcing me to install docs when i don’t want them. 
    greetings, eMBee.

  2. shaban said says:


  3. I agree with most of your points, except the one about screen shots. I understand the appeal of screen shots, but they present a maintainability challenge. Even minor wording differences between the application and the document can cause confusion. It gets worse when someone comes along and wants to translate your document into another language. Now they have to reproduce all of your screen shots, too. For small documents that might not be so bad, but for larger documents it begins to be unwieldy.

  4. The absolute worst docs I’ve ever used are the ones for git. I love git, but understanding some of its concepts for new users can be hard. Especially when the docs seem to be written for computer scientists, and the examples have no bearing on reality (when there actually are examples)! They lack simple overviews of each feature. They really do their project and their users a disservice. Git is such a great tool, and im convinced it would be much more widely used if its docs weren’t so preventitivly cryptic.

    • Agreed, I learn git from coworkers and friends. Even sometimes need to look at doc , I still read someone’s blog but the manual.

  5. I agree with most you’re writing. What I want to see in a manual is as follows: 
    – What has been changed since last version 
    – Overview (what is it, what does it, design objectives) 
    – Guided tour (a sample session) 
    – Installation manual 
    – Tutorial (basics and advanced) 
    – Reference guide 
    (glossary (if coding tool), error guide, standards compliance, etc.) 
    – Development guide (if you want to change the code; overview, API, programming examples) 
    I learned a lot from the standard Unix documentation and the Borland manuals in this regard. And true, most manuals are severely lacking. I’m still waiting for a GCC development guide, for example. 
    Someone who who a 550 page manual for his FOSS project.

  6. Amen. 
    The documentation of the Spring framework is the way to go for me: Extensive user guide with lots of real-life examples and very concise JavaDoc. Of course, Spring being a framework, there are no screenshots.

  7. This is spot on. Most of the problems with open source software related to the fact that is ALWAYS a work in progress. Most of these guys just wont let well enough alone, so the documentation is always months behind and therefore confusing at best to downright worthless at worst.

  8. most developers hate preparing documentations and this shows in the quality of the information.  
    whenever we work on a documentation for a customer, we also try to provide PPTs with the most important functionalities, so that the flow is easily understood. also, this helps us to diminish the quantity of text, as images tend to provide more info.

  9. I have never read the .doc for any software I have used. I don’t feel like I am any kind of exception.

  10. About point 7, “Best of all, do both!” 
    Please don’t! Write only an explanation for a term or concept the first time you talk about it and then always link to it. 
    Try to never repeat explanations in your docs, it’s boring to read the same over and over. 
    And worst, eventually you’ll forget to update one of those explanations and your docs will become confusing. 

  11. Documentation is one of the main reason why I now use mosty FreeBSD and OpenBSD instead of linux, except for PBX. 
    Most BSD doc are awsome. I remember trying to configure something in debian network config script and don’t find it anyware except for a blog. The network script online man page was several years old and missing it.

  12. John Lenihan says:

    All of these failures are always profoundly frustrating! Most result from functional illiteracy and massive developer egos which can’t conceive of any reason why users don’t bow deeply before the originating genius! 
    Example: In trying to configure the Ekiga softphone , I never found an answer to the difference between menu entries User and User Authentication, despite a manual with thousands of words, and huge web searches. Using the same value for one or both caused install failure. Stupid and worthless program!

  13. Good article. Agree with all of it. I’d add: requiring people to register and or jump through hoops to get information.

  14. Vincent Manis says:

    This is a very good article, and software developers of all backgrounds ought to heed it carefully. I would add a key point: provide a linear path through the documentation for those learning the software.  
    Many outstanding systems are provided with a collection of HTML files; with proper linking, it’s possible to use this approach to produce excellent reference documentation, but it’s very difficult to navigate a maze of web pages if you’re trying to learn the system in the first place. Before knowing that the `frob’ command creates a new blivet, I kind of have to know that the program manipulates blivets, that I must enable blivets in the configuration file, and that I can list the active blivets by using the `wazoo’ command.  
    Some projects respond to this by creating separate tutorials that never get updated, so the tutorial might never mention blivets at all, even though they have been in the past four major version releases.  
    I still keep an old version of Adobe Acrobat (running on an XP virtual machine, they are the only commercial programs I own). When I get one of these tangled webs of documentation, I fire up Acrobat and make a PDF out of it. I used to print it out; now I can load it onto my tablet, and make a pass through the material to get a sense of how it all works. 
    Although the documents are generally a bit on the verbose side, I do have to commend the GNU project for creating documents that can be read as a manual but also used effectively as reference material. There are many other projects that have done the same.

  15. I am a 85 yo user and from my position you have pointed out about 95 percent of the new user problems – thank you! 

  16. Developers need to remember that what is obvious to them may not be to the user.  
    I would suggest software developers who find documentation a chore, please put out requests for help.  
    In regard to Linux distributions, I know some have okay help docs, but others seem to leave potential users high and dry, referring them to forums and Internet searches. And worse, many often don’t include off line help for common issues or known issues and their fixes, this is especially important for connectivity issues. Needing to rely on a smartphone or another computer connected to the net to find answers is a horrible experience.

  17. I hope many people within software projects are going to read this insightful article.  
    Here’s an acid test for your user guide or other documentation: Give it to an inexperienced user and observe how well he can perform by following the doc. Where the user stumbles, you have to make the guide better. Simple. Effective.

  18. Don’t describe your project in relation to another project, such as “NumberDoodle is like BongoCalc, but better!” That’s no help to someone who’s never heard of BongoCalc. 
    Irony: the above written by the domain owner of betterthangrep.com.

  19. Most open source projects need you to be ready to spend significant amount of time trying things out by yourself. And god help you if you end up discovering a bug in it. If its popular, you might find some support in a forum somewhere but if you are using it for an enterprise application, you better be prepared to explain why you used it, to your management.

  20. Try reading Apache Tomcat documentation, worst documentation ever

  21. What I really hate about this blog is that blue menu thingie on the left side, which is a nuisance and distracts from the rather good content. 😉

    • @jhermann:disqus, thanks for the feedback. We’re just doing a test run of this recommendations widget to see if it is helpful or intrusive. Thanks for letting us know your opinion. (We’re also glad to hear you enjoyed the content.)

  22. Thanks for that great article! I’d basically agree with 3 but I would consider it the least important. It surely is nice to have any documentation installed. But I’d say, having it online only isn’t a major hassle for lots of users. Online has many advantages over offline docs (up to date, commenting, peer edit etc.) and nowadays lots of people hold their smartphone like they used to hold the documentation in the past 😉
    I’d say, if the online documentation doesn’t provide an easy way to be packaged, 99% of all projects got better things to do than achieving that. (Let it be the improvement of the online docs).
    Also, there is a market for printed 3rd party literature on FOSS projects that is used by people who prefer physical/offline documentation.

  23. When you make a website, all your web pages are
    served from the server residing somewhere on the internet.
    Dream – Host is an online hosting company that offers products like domain registration,
    shared website hosting, Virtual Private Server (VPS), and public cloud services.
    You can be diligent about protecting your website but someone
    else on the same server might not be.

  24. Appreciate this post. Let mee try it out.

  25. Hello to every one, it’s truly a good for me to visit this web page, it consists of important Information.

  26. What i don’t realize is in reality how you are now not really a lot more
    well-favored than you may be now. You are very intelligent.
    You know therefore significantly in terms of this subject,
    produced me in my opinion imagine it from a lot of varied angles.

    Its likke women and men are not involved except it’s something to do with Girl gaga!
    Your individual stuffs excellent. Always take
    cwre of it up!

  27. This paragraph is genuinely a nice one it assists new net
    people, who are wishing in favor of blogging.

  28. This piece of writing is genuinely a nice one it helps new the web
    visitors, who are wishing for blogging.

  29. How about angry and irritable maintainers?

    Dear Maintainer. Accept the price of success. People will use your software and ask you loads of questions about it.

  30. You’ve raised a pretty valid point about the quality of help available for the open source resources. But it is mostly due to lack of allocated resources for structuring and proofing documentation. Also, the focus of the open source creator is on the end product and ensuring the completeness of its features. Being a consumer of such resources, it is also our responsibility to help improve the documentation by timely informing the source.

  31. Truly no matter if someone doesn’t be aware of after that its up to other users that they will
    assist, so here it happens.

  32. Salaam,

    Gasping at your brilliance! Thanks a tonne for sharing all that content. <a href="https://asha24.com/blog/workday-tutorials-training-certification-and-interview-qa"Can’t stop reading. Honestly!

    We have locations defined as part of legal entity configuration, these are marked as “Legal address” in HRMS Location form.What is the downstream/upstream impact if I choose a different ship-to location (previously defined) for this location in the shipping tab. I also wish to uncheck Receiving Site. I do not want to use this location in a Purchase Order, but don’t want to cause any issues in any other process as part of the configuration.
    Excellent tutorials – very easy to understand with all the details. I hope you will continue to provide more such tutorials.


Speak Your Mind