A lot of technical writers focus on the style of the text. For example, I've seen tech writers have long discussions (more than 3 hour-long meetings) about title capitalization in our organization.
I wish tech writers would channel their focus elsewhere:
- If you haven't got it working, don't document it. Getting information by talking to SWEs is not enough. Your docs will lack depth and preciseness, and it will show. A lot.
- Show fully qualified imports in code samples. If you don't know what that is, ask SWEs. Given a computer with installed dependencies (such as Maven, Pip, npm, etc.), even your (grand)mother should be able to follow the docs.
- Create automated testsuide for your code samples.
- Mandate that lack of docs be a blocker for a release.
When there are docs written in perfect style and English, yet none of the code samples works, or when I have to spend 30 minutes finishing the "obvious" parts of the code to get it to work, that's when I get really frustrated.
The article seems to be deeply focused on the style. I can't remember when I got frustrated with documentation because the style was bad, or because the sentences were too long.
A fair observation. I do think there's some value to concerns over style, but far too many tech writers give them undue prominence.
I couldn't champion your bulleted points harder. Honestly, if I had a say in it, any tech writer working on developer documentation should be able to code/build/perform/use whatever task or api they're documenting without the help of an engineer. They should be able to read and understand code, and they should be willing to do so, even more so than the average engineer using some library code. They should be capable of understanding a library deeply and synthesizing and summarizing that understanding for users of the library.
Just as tech writers in the medical field need to know quite a bit about medicine, tech writers in software should be required to know quite a bit about engineering software. There's plenty that do, but there's also quite a few that don't--I think it's a side effect of the field's relative youth.
I really wish more writers would code and do exactly the things you mention, such as implement robust code sample testing, etc.
The ideal technical writer, at least for software engineering documentation, is a hybrid of a traditional technical writer and a software engineer[^1].
[^1]: This excludes user-facing software documentation, which is essentially a totally different field than those that deal with writing for developer or engineer audiences.
The problem with your perspective is an economic one. If you have the chops to be a hybrid TW/SWE, you will never take a tech writing job (or get out of tech writing as soon as you can) because tech writers simply make substantially less than engineers. Everyone acknowledges that tech writing is vital but until the pay story reflects that the status quo will persist.
> ...any tech writer working on developer documentation should be able to code/build/perform/use whatever task or api they're documenting without the help of an engineer. They should be able to read and understand code, and they should be willing to do so, even more so than the average engineer using some library code. They should be capable of understanding a library deeply and synthesizing and summarizing that understanding for users of the library.
This is what made my friend Caroline Rose such a great technical writer. She not only can explain things in clear and simple terms, but she was also an accomplished programmer before branching into technical writing.
Here is a story about the documentation that became Inside Macintosh:
> I sat down to meet with Caroline for the first time, and she couldn't have been more different than the previous writer. As soon as I began to explain the first routine, she started bombarding me with questions. She didn't mind admitting it when she didn't understand something, and she wouldn't stop badgering me until she comprehended every nuance. She began to ask me questions that I didn't know the answers to, like what happened when certain parameters were invalid. I had to keep the source code open on the screen of my Lisa when I met with her, so I could figure out the answers to her questions while she was there.
> Pretty soon, I figured out that if Caroline had trouble understanding something, it probably meant that the design was flawed. On a number of occasions, I told her to come back tomorrow after she asked a penetrating question, and revised the API to fix the flaw that she had pointed out. I began to imagine her questions when I was coding something new, which made me work harder to get things clearer before I went over them with her.
One of my favorite examples from _Inside Macintosh_ is the discussion about the mathematical foundation of QuickDraw. Modern graphic programmers all know that a (0,0,0,0) rectangle is an empty rectangle. But why is that? Why isn't it a single pixel?
The explanation on pages I-138 through I-141 is what made this concept make sense to me. Here's a PDF of _Inside Macintosh_:
Besides that section, there are all kinds of examples of how to explain a complicated concept in simple and understandable terms. It's worth a quick skim of the entire book to see what good documentation looks like.
The issue with that though is that if someone could read/write code, they would probably choose to be a SWE instead of a tech writer.
I'm not sure why this is, but I suspect that it's an combination of higher pay and an interest in writing code over writing documentation.
In my experience I would say that a good SWE will attempt to write good documentation, as having to answer questions in the future about how stuff works over and over takes more time than just documenting things properly the first time.
How does one become a technical writer? I'm a software engineer, and e.g. I recently applied to Google for both a tech writer position and a software engineer position, and they interviewed me and offered me a job for the software engineer role but they gave me a form rejection for the tech writer role. I worry I'm stuck as a software engineer because that's what's on my resume.
Also, document the data model more than how the code works. Famous Linus Torvalds quote: "Bad programmers worry about the code. Good programmers worry about data structures and their relationships."
I don't see where he says anything about style. Could you point out to me what he says that makes you think that?
He does say to use bullet lists, but I think this is more about structuring the content and breaking ideas into steps than it is about style or formatting.
His points all seem to focus on providing clarity and useful content to the reader and user.
> Could you point out to me what he says that makes you think that?
I'd consider all of the following to be mostly style choices:
* Avoid Meta Writing
* Minimalist Instruction
* The "Constructivism" section, e.g.--Engage in an active dialog, invite users to act “Please try”, “See what happens”, “Explore yourself”
(The above is now what I'd estimate to be like 80% of the article)
The reason why I'd consider that a style is that if you buy a style guide, e.g. the IBM style guide, it would include similar instructions, e.g. "avoid passive voice" or "be more succinct" (though at a more of an implementation level).
It's possible you'd have a better name for it than 'style'. I'd accept that. My original point stands though; most TWs I've seen concern themselves with the letters of the documentation, and omit what I'd consider the most useful about documentation, that is the deep technical details such as maintaining code samples, or writing test suite for the documentation.
Even a computer without installed dependencies should be able to follow the docs. Some don’t list what dependencies they need at the beginning, which could save a lot of time and error messages.
Totally agree with your points about code correctness. That's the most important thing. I spent probably a year making systems/tools/processes to ensure that every code snippet in our documentation is imported unmodified from a GitHub repo with CI attached and that people reading the docs can see that. It's been a huge win.
When I first saw the title I thought to myself, “shit, another random, possibly misleading post about technical writing?” but on quick review this seems to be a well-researched overview of the field. If you were to ask a practicing technical writer to present a summary like this, it wouldn’t differ much.
If the author is here, I’m curious if they took that definition of technical writing from my post? [1] To the best of my memory I feel like I came up with that definition on my own but also wouldn’t be surprised if I subconsciously took it from somewhere else.
I am here, and I have read several of your posts, thanks for wonderful resource. So yes, it is quite possible I took the definition from you, I collected the various notes over time, so I'm not always positive where they come from.
> In this time, I’ve read various resources on technical writing and documentation. These are my notes, both to help me remember later, but also as a tool to help me think about writing now.
and later, in an apparent contradiction ...
> You don’t need to tell the reader what you’re going to tell them. Just tell them.
I think this point could be fleshed out better.
Any piece of nonfiction writing, technical or otherwise, that doesn't give me a clear idea about its content may not pass my filter. I see this the most in long-form journalism. But it also pops up when authors focus on themselves, their struggles, and their aspirations in the first couple of paragraphs.
Telling the reader what they're about to read in the first paragraph of a technical piece is a courtesy. People are busy, and they don't have time to wade through a bunch of throat-clearing at the start of a piece. They'll just close the tab and move on.
This is a good counterpoint. After all, academic journals have abstracts for a reason--when people need to get something done, a synopsis quickly helps them make a decision as to whether or not they actually found the appropriate resource.
It goes both ways. I think this aspect of the craft is a bit more of an art than a science. At certain points it's appropriate to signal what you're about to discuss, at others, it's just noise.
In general, anything that functions similarly to an academic abstract is probably useful (saves the reader from potentially wasting time, gives an overview of the general subject matter and idea without delving into details, takes about 1 minute maximum to read).
Abstract or "executive summary" is the paragraph right under the title that does this in everything I wrote for the exact purpose to give someone the opportunity to see whether this is his/her go-to document for a purpose or note. I always like to add/read the ToC, that is also a quick read to see if this doc is what I am/people are looking for.
I've also noticed that for some reason people get sloppy when writing posts about technical writing and forget to apply the principles of technical writing to that post. I avoid this bad habit by forcing myself to provide a summary for every post I create, e.g. https://kayce.basqu.es/blog/FAQs/
>I've also noticed that for some reason people get sloppy when writing posts about technical writing and forget to apply the principles of technical writing to that post.
Well, technically meta-posts about technical writing are not technical writing.
Any piece about writing will violate its own rules. I think this could be a law ;)
This is because writing is hard. It is just like coding, in the sense that I can revise it three times and still find ways to make it better the fourth time, or the tenth time.
So no writing is perfect, but I think I can often tell when a writer has read the right books (like On Writing Well). Every now and then an article will jump out to me as being 10 times more lucid and concise than the norm --- even though it's still imperfect.
Also I think the rules are not black and white. There is some judgment needed at every turn. Sometimes you will want to bend one rule to preserve another, and people with different tastes will either praise you or blame you for how you make the call.
I like the rule of three. Tell the reader what you are going to tell and why. Tell them in detail. Then in the conclusion summarize the main takeaway from what you just told them.
I've never understood the pre-summarization technique. I've had lecturers constantly state what they intend to say so much that its a distraction. When I see this style of writing I glaze over as well.
I don't think it's an apparent contradiction. The first quote is a meta description of the whole article. People would need to know what it's about to read it.
In actual technical documentation, on the other hand, people already know "this is the documentation of program X". And in the various chapters, you can just go ahead and give them a good title, and then go say what it is to say - without first announcing what you'll say.
I always try to understand the readers before writing a word. Why they are reading. And what they expect to be able to do better as a result of reading, or what decision they might be making.
If an author can't answer these questions, they will likely write something that is perceived as irrelevant, wrong, or simply 'meh'. This is especially true when writing 'technical reports' as opposed to documentation.
"Know thy audience" and "know thy audience's purpose" are the cardinal rules of technical writing, as mentioned in Marcus's post. Often times there are limits to how much I can know about my audience. In that case it's especially helpful to put an "Assumptions" or "Intended audience" section at the top of a doc.
+1 on that "thy audience", and the audience is not just the end-users. It is the 10 extra layers/people that will get to carry an opinion (and whose opinion matters) by glancing a document for 1min or less.
"Don’t tell the reader something is extremely complex; likewise don’t tell the reader something is simple. The reader will make up their own mind about a concept being easy or hard."
I think this is great advice, and applied to teaching. I can't tell you the number of times a professor has said "and then obviously it follows that" or "it's easy to see that". Nearly every time that was uttered in a lecture I wanted to hurl my textbook at the board.
Looking for alternatives? "Is it easy?" or "It might be obvious" work fine. Maybe use those moments as opportunities to check in with your audience. In the written context, maybe offer end/footnotes explaining why it might be obvious/easy.
Related: don't say someting is interesting. For example, "It's interesting that . . . " I see this preface more often than "simple" or "complex." Also there is a type of person for whom your introduction will come across as a dare to find it uninteresting.
Most documentation is so bad and so often out-of-date that I can answer my question faster by reading the source (or tests).
Everything is so extremely verbose. I don't want to read a 7 paragraph essay about how this or that feature was developed and evolved over time; I just need a one line hint and maybe an example. Believe it or not, your software is not special. We've seen something like it before and get the gist already. I like the style of man pages, where everything is on one page and organized by the flag, i.e. actual usage (puts a hard limit on meandering bullshittery).
You are in the minority, mate. Man pages are walls of text that cover too much detail in too little space. A lot of them lack examples.
Most people will take a simple task-oriented tutorial over an all-encompassing man page. That’s something that Microsoft understood very well back in the ‘90s, providing copious amounts of examples and tutorial on top of basic reference material. It’s part of the reason they saw orders of magnitude more adoption for their stack.
> Most documentation is so bad and so often out-of-date that I can answer my question faster by reading the source
Especially within a company. Once place I worked, a few teams liked to over-document, though it wasn't enforced org-wide. The docs were usually not very helpful, outdated, and yes, I could just read the code faster. I wish they put more effort into better factored code and better names.
Keep documentation close to the source to increase the odds it gets updated. Prefer READMEs to Wikis, and use source comments to generate docs rather than word/google docs. Some doc tools even test examples for you!
It's not perfect, and good documentation is an art, but if you care about your software and your users you'll try to communicate clearly.
I totally agree with this. You have to teach critical concepts a couple different ways in case one gets skimmed over or misunderstood. We rely on visuals for this--the text should stand independent of the photo / diagram, and vice versa.
Another hugely important reason to support error recovery: this is how many people find documentation. They see an error when using an application (or writing code) and they copy-paste the error into a search engine. If you provide that error text verbatim in your documentation, you will capture that search engine traffic.
What's needed is something more like peer review, where the "tech writer's" job is to ask hard question from the user POV during development - for varying values of "user", which would include customer, API consumer, and so on.
This wouldn't be code review, because it's about features, interfaces, and expected behaviours, not the fine details of implementation. And it wouldn't be a PM job, because it would be peer feedback in the form of dialog, not management in the form of requirements and specs.
And it should be continuous, so the doc base evolves in parallel with the code base.
And it should not be an afterthought.
This job does not exist yet. And it should, because virtually all orgs have a problem where individual developers become repositories of The Important Knowledge. If they leave it becomes hard to characterise system behaviours and architectures well enough to understand how they work well enough to replace/improve/reverse engineer them.
So IMO there's a niche for people who are competent developers, competent listeners, and competent communicators, who can pull it all together, stay on top of changes, and leave a readable record of the current state of institutional knowledge about projects, practices, stacks, and interfaces.
Readit News
I wish tech writers would channel their focus elsewhere:
- If you haven't got it working, don't document it. Getting information by talking to SWEs is not enough. Your docs will lack depth and preciseness, and it will show. A lot.
- Show fully qualified imports in code samples. If you don't know what that is, ask SWEs. Given a computer with installed dependencies (such as Maven, Pip, npm, etc.), even your (grand)mother should be able to follow the docs.
- Create automated testsuide for your code samples.
- Mandate that lack of docs be a blocker for a release.
When there are docs written in perfect style and English, yet none of the code samples works, or when I have to spend 30 minutes finishing the "obvious" parts of the code to get it to work, that's when I get really frustrated.
The article seems to be deeply focused on the style. I can't remember when I got frustrated with documentation because the style was bad, or because the sentences were too long.
I couldn't champion your bulleted points harder. Honestly, if I had a say in it, any tech writer working on developer documentation should be able to code/build/perform/use whatever task or api they're documenting without the help of an engineer. They should be able to read and understand code, and they should be willing to do so, even more so than the average engineer using some library code. They should be capable of understanding a library deeply and synthesizing and summarizing that understanding for users of the library.
Just as tech writers in the medical field need to know quite a bit about medicine, tech writers in software should be required to know quite a bit about engineering software. There's plenty that do, but there's also quite a few that don't--I think it's a side effect of the field's relative youth.
I really wish more writers would code and do exactly the things you mention, such as implement robust code sample testing, etc.
The ideal technical writer, at least for software engineering documentation, is a hybrid of a traditional technical writer and a software engineer[^1].
[^1]: This excludes user-facing software documentation, which is essentially a totally different field than those that deal with writing for developer or engineer audiences.
This is what made my friend Caroline Rose such a great technical writer. She not only can explain things in clear and simple terms, but she was also an accomplished programmer before branching into technical writing.
Here is a story about the documentation that became Inside Macintosh:
> I sat down to meet with Caroline for the first time, and she couldn't have been more different than the previous writer. As soon as I began to explain the first routine, she started bombarding me with questions. She didn't mind admitting it when she didn't understand something, and she wouldn't stop badgering me until she comprehended every nuance. She began to ask me questions that I didn't know the answers to, like what happened when certain parameters were invalid. I had to keep the source code open on the screen of my Lisa when I met with her, so I could figure out the answers to her questions while she was there.
> Pretty soon, I figured out that if Caroline had trouble understanding something, it probably meant that the design was flawed. On a number of occasions, I told her to come back tomorrow after she asked a penetrating question, and revised the API to fix the flaw that she had pointed out. I began to imagine her questions when I was coding something new, which made me work harder to get things clearer before I went over them with her.
https://www.folklore.org/StoryView.py?story=Inside_Macintosh...
One of my favorite examples from _Inside Macintosh_ is the discussion about the mathematical foundation of QuickDraw. Modern graphic programmers all know that a (0,0,0,0) rectangle is an empty rectangle. But why is that? Why isn't it a single pixel?
The explanation on pages I-138 through I-141 is what made this concept make sense to me. Here's a PDF of _Inside Macintosh_:
http://www.weihenstephan.org/~michaste/pagetable/mac/Inside_...
Besides that section, there are all kinds of examples of how to explain a complicated concept in simple and understandable terms. It's worth a quick skim of the entire book to see what good documentation looks like.
I'm not sure why this is, but I suspect that it's an combination of higher pay and an interest in writing code over writing documentation.
In my experience I would say that a good SWE will attempt to write good documentation, as having to answer questions in the future about how stuff works over and over takes more time than just documenting things properly the first time.
He does say to use bullet lists, but I think this is more about structuring the content and breaking ideas into steps than it is about style or formatting.
His points all seem to focus on providing clarity and useful content to the reader and user.
I'd consider all of the following to be mostly style choices:
* Avoid Meta Writing
* Minimalist Instruction
* The "Constructivism" section, e.g.--Engage in an active dialog, invite users to act “Please try”, “See what happens”, “Explore yourself”
(The above is now what I'd estimate to be like 80% of the article)
The reason why I'd consider that a style is that if you buy a style guide, e.g. the IBM style guide, it would include similar instructions, e.g. "avoid passive voice" or "be more succinct" (though at a more of an implementation level).
It's possible you'd have a better name for it than 'style'. I'd accept that. My original point stands though; most TWs I've seen concern themselves with the letters of the documentation, and omit what I'd consider the most useful about documentation, that is the deep technical details such as maintaining code samples, or writing test suite for the documentation.
If the author is here, I’m curious if they took that definition of technical writing from my post? [1] To the best of my memory I feel like I came up with that definition on my own but also wouldn’t be surprised if I subconsciously took it from somewhere else.
[1] https://kayce.basqu.es/blog/metrics/
An aside: did you redesign your website, Kayce? I recall it looking quite different in the past. It looks great! I like the aesthetic.
and later, in an apparent contradiction ...
> You don’t need to tell the reader what you’re going to tell them. Just tell them.
I think this point could be fleshed out better.
Any piece of nonfiction writing, technical or otherwise, that doesn't give me a clear idea about its content may not pass my filter. I see this the most in long-form journalism. But it also pops up when authors focus on themselves, their struggles, and their aspirations in the first couple of paragraphs.
Telling the reader what they're about to read in the first paragraph of a technical piece is a courtesy. People are busy, and they don't have time to wade through a bunch of throat-clearing at the start of a piece. They'll just close the tab and move on.
It goes both ways. I think this aspect of the craft is a bit more of an art than a science. At certain points it's appropriate to signal what you're about to discuss, at others, it's just noise.
In general, anything that functions similarly to an academic abstract is probably useful (saves the reader from potentially wasting time, gives an overview of the general subject matter and idea without delving into details, takes about 1 minute maximum to read).
Well, technically meta-posts about technical writing are not technical writing.
This is because writing is hard. It is just like coding, in the sense that I can revise it three times and still find ways to make it better the fourth time, or the tenth time.
So no writing is perfect, but I think I can often tell when a writer has read the right books (like On Writing Well). Every now and then an article will jump out to me as being 10 times more lucid and concise than the norm --- even though it's still imperfect.
Also I think the rules are not black and white. There is some judgment needed at every turn. Sometimes you will want to bend one rule to preserve another, and people with different tastes will either praise you or blame you for how you make the call.
In actual technical documentation, on the other hand, people already know "this is the documentation of program X". And in the various chapters, you can just go ahead and give them a good title, and then go say what it is to say - without first announcing what you'll say.
If an author can't answer these questions, they will likely write something that is perceived as irrelevant, wrong, or simply 'meh'. This is especially true when writing 'technical reports' as opposed to documentation.
I think this is great advice, and applied to teaching. I can't tell you the number of times a professor has said "and then obviously it follows that" or "it's easy to see that". Nearly every time that was uttered in a lecture I wanted to hurl my textbook at the board.
Looking for alternatives? "Is it easy?" or "It might be obvious" work fine. Maybe use those moments as opportunities to check in with your audience. In the written context, maybe offer end/footnotes explaining why it might be obvious/easy.
Deleted Comment
Roughly: Don't write item/place/npc descriptions that tell the player ("you") anything they should think or feel.
Easier said than done, though.
Everything is so extremely verbose. I don't want to read a 7 paragraph essay about how this or that feature was developed and evolved over time; I just need a one line hint and maybe an example. Believe it or not, your software is not special. We've seen something like it before and get the gist already. I like the style of man pages, where everything is on one page and organized by the flag, i.e. actual usage (puts a hard limit on meandering bullshittery).
You are in the minority, mate. Man pages are walls of text that cover too much detail in too little space. A lot of them lack examples.
Most people will take a simple task-oriented tutorial over an all-encompassing man page. That’s something that Microsoft understood very well back in the ‘90s, providing copious amounts of examples and tutorial on top of basic reference material. It’s part of the reason they saw orders of magnitude more adoption for their stack.
Especially within a company. Once place I worked, a few teams liked to over-document, though it wasn't enforced org-wide. The docs were usually not very helpful, outdated, and yes, I could just read the code faster. I wish they put more effort into better factored code and better names.
It's not perfect, and good documentation is an art, but if you care about your software and your users you'll try to communicate clearly.
> Support error recognition and recovery.
I totally agree with this. You have to teach critical concepts a couple different ways in case one gets skimmed over or misunderstood. We rely on visuals for this--the text should stand independent of the photo / diagram, and vice versa.
I wrote a handbook on technical writing summing up what we've learned over many years of writing iFixit repair guides. https://help.dozuki.com/Tech_Writing/chapter/0
What's needed is something more like peer review, where the "tech writer's" job is to ask hard question from the user POV during development - for varying values of "user", which would include customer, API consumer, and so on.
This wouldn't be code review, because it's about features, interfaces, and expected behaviours, not the fine details of implementation. And it wouldn't be a PM job, because it would be peer feedback in the form of dialog, not management in the form of requirements and specs.
And it should be continuous, so the doc base evolves in parallel with the code base.
And it should not be an afterthought.
This job does not exist yet. And it should, because virtually all orgs have a problem where individual developers become repositories of The Important Knowledge. If they leave it becomes hard to characterise system behaviours and architectures well enough to understand how they work well enough to replace/improve/reverse engineer them.
So IMO there's a niche for people who are competent developers, competent listeners, and competent communicators, who can pull it all together, stay on top of changes, and leave a readable record of the current state of institutional knowledge about projects, practices, stacks, and interfaces.