March 14th, 2016
I was asked to give a talk at the Stevens Institute of Technology Computer Science Club on the topic of things that are not taught in school, but which CS students should know. I asked around a bit and got a lot of really interesting and useful feedback, leading to a long laundry list of possible topics. In the interest of time, I had to cut out a whole lot and still ended up with over 90 minutes of material. The end result is below.
The slides for this talk are also availablehere as well as on slideshare .
This is going to be long, and it will cover a lot of ground, and it will be completely subjective and it will be incomplete. A lot of this will also be obvious. Or perhaps only obvious in hindsight , because of course once you hear it, or you’ve experienced it, it’s trivial to say that you’ve known this all along.
I will try to focus on things that I believe they do not teach you or do not stress enough in your education. Some of these things used to be taught here at Stevens; some of these things they have started to teach. Some of them they won’t ever teach, because they do not fit into the academic goals of a CS curriculum nor the general concept of a university level education.
There is a discrepancy between what the industry requires or expects from academia, and what schools and universities provide . There are some things that universities are not prepared or willing to teach students simply because it does not fall within their mission, and whether or not that is Right and Good is a topic for another day. The industry, on the other hand, makes the mistake of expecting universities to churn out "programmers", ready to join the big machine and crank out code. But programming is but one aspect of all the many things that relate to "Computer Science".
The list of things I’ll talk about is necessarily incomplete. I also cannot possibly actually teach you all the Things They Don’t Teach You In School. But I can at least tell you what some of them are, so that you can begin to research them, to focus on them, to find out more, so that at least now you know some of the things you don’t know.
And that is already one of the main lessons: Know your Unknowns. Know what you don’t know. Now I’d be the last person to look favorably upon Donald Rumsfeld , but unfortunately it is his name that has become associated with this concept:
"There are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don’t know we don’t know."
Or, perhaps phrased as the Socratic paradox : I know that I know nothing .
This is actually important and not as confusing as it may sound. And it has some interesting implications on your confidence. You may have heard of the terms " Dunning-Kruger effect " and " impostor syndrome ", which are basically two sides of the same coin:
The fallacy here is that the more you know, the more you are aware of all the things you don’t know, thus losing confidence and feeling like you know so much less than you actually do.
Undergrads — by and large, anyway — know that they don’t know much. Once they graduate, however, something weird happens: with this degree in hand, they feel like they know a whole lot. Why, they even implemented Huffman compression once! They’ve written literally hundreds of lines of Java and even some C++, so yeah, they’re "expert" or at least "strong" programmers. Right?
At this point, the ratio of the things they know they don’t know versus the things they actually know is small. I.e., they think they know more than they don’t know, thus developing strong confidence in their abilities.
As time goes by and you learn more and more, you also learn that there are a lot more things out there that you don’t know. So even though the set of things you do know has increased, the ratio of the things you know you don’t know versus those you do know has shifted. As a result, you feel like you know (comparatively) less.
We could have a full talk on just this one syndrome, but suffice it to say that it’s important to have a reasonably realistic understanding of what you know and what you don’t know.
Dunning-Kruger and Impostor Syndrome are entirely fascinating and have profound impact on how you approach your work, your peers. Which gets me to my next point:
Everything you do in this field ultimately boils down to solving people problems, to understanding human beings and their motivations. Somehow people in the tech world pretend that they can decouple the human component from the technical requirements or the general problem space, and that never works out.
We’ll get back to this underlying concept throughout today’s talk.
The reason that you attend college or any higher education is not to learn individual skills. You are here to learn concepts . To learn how to learn. You do not get a CS degree to become a programmer — you can learn how to program yourself, in less than 24 hours, no less. ( Right? )
Yes, this is how programming is done, by and large. But note that what happens in the so-called "real world" is not a good idea in school: by doing this, you are defeating the point of the exercise.
One part of the exercise is for you to do the work . When we ask you to write a program to sort an input set, then we are doing that not because we need a program to sort an input set. We are giving the assignment so that you write it, so that you get the experience of implementing the algorithm. Copy and paste driven development will rob you of this, the main objective of the assignment.
You may have heard that it takes around 10K hours to become an expert in just about anything. So you better get a head start:
If you want to get a job in that so-called real world, you better have some programming experience to show for. Academia will teach you concepts, which are important, but the overwhelming majority of jobs out there do not require an in-depth understanding of Computer Science and instead focus more on practical programming experience.
It’s true that understanding the CS concepts we’re teaching you here at Stevens will make you a better programmer (if that’s what you want to do), but the theory is necessary, not sufficient , to becoming a good programmer, a good software engineer, or whatever it is you end up doing with this degree.
There are some things you can only learn by doing; some lessons you cannot learn in the abstract.
In CS, you learn about Operating System concepts such as virtual memory or filesystems, or perhaps distributed computing algorithms and things like the Dining Philosophers or the Thundering Herd problem. But unless you have experienced them, you do not fully understand them. Unless you have actually implemented a web server or an http client, you do not understand all the details of HTTP.
This also means that the first time you try to solve a problem, you necessarily are going to produce a a flawed, incomplete solution. As you implement your tool or program or algorithm, you are learning the many details that weren’t even on your horizon when you began, so your solution is necessarily… crappy. Throw it out!
Be aware that your first solution is nothing but a prototype. A first attempt at solving the problem, used to outline the general solution and exploring the problem space. If you’re fine with that, then it’ll be easy to junk it and begin work on the actual solution.
(Sidenote: beware " Second System Effect " and feature creep.)
You know why prototypes are so hard? On the one hand, we are unwilling to throw away what we worked so hard on. Imagine: you spent all this time in writing a program that does what you think is necessary, and you learned all the intricate details of the problem at hand, and now you’re supposed to just throw all that away?
But if you do that, you are not actually throwing away all that much. You can much more easily build a new version once you’ve done it the first time. Have you ever worked hard on a homework assignment and then accidentally deleted all or a big chunk of your work? Sure, it’s frustrating, but what took you maybe a week to write the first time will take you only a few hours the second time, because now you much better understand the problem space.
So don’t be afraid to throw away your prototype and begin work anew. Otherwise, your risk that your prototype, intended to be a temporary fix, becomes a permanent solution.
One area where this quickly becomes problematic is when you never intended the program in question to be the be-all-end-all solution. You just wanted to fix something real quick. Maybe you were paged in the middle of the night and you wanted to just solve the immediate problem and go back to bed and tomorrow morning you’ll fix it the "right way", right?
There is nothing more permanent than a temporary solution. Once you have put in place a "quick fix", you are going to forget about solving the problem the right way; something else will come up to distract you, something more important (because your quick fix works, and so this particular part of your world is not on fire any more), or somebody else has already built something that depends on your "quick fix"’s particular behaviour.
You may have heard this before: if you keep trying to build the perfect system, to address all use cases, to meet all requirements, you’ll never ship your product. It will never be finished.
" Worse is Better " (also called the "New Jersey style") is an important essay on software development that everybody in this field should have read. It’s main point is that software does not primarily need to be "correct" to be "better". Quality does not increase with functionality, and it is often preferable to focus on delivering a "worse" product (fewer features, incomplete implementation, possibly even inconsistent design) that is simpler. (There are some parallels here to the idea of the "minimum viable product".)
Worse is better, but there’s an interesting corollary:
If you have a solution that kind of "stops the bleeding"; that allows you to focus on something else; that takes away the urgency of the problem, then you are much less likely to come back to address the core problem.
A lot of software development, programming, operations etc. is done in a firefighting mode. If what you have "kind of" works, then few people will see the necessity to go back and redo the work, to do it the right way.
So even if it may seem paradoxical, sometimes it’s better to object to a solution that addresses the immediate needs (ie the "quick fix", the temporary solution) in favor of solving the whole problem in the correct way.
One trick to balance "worse is better" and "perfect is the enemy of the good" is to focus on Simplicity .
Simpler code has fewer bugs. Simpler code is more readable. Simpler programs are easier to understand, easier to use. This is important in how you design your programs or APIs as well as how you write them.
The software you write should do one thing and do that thing well.
Follow the Unix philosophy . In Unix, your tools interface with one another via text streams going through stdin and stdout. This is a simple interface, and an example of "Worse is Better" — designing pipes as passing objects would require a complex design and reduce the flexibility of those implementing new tools you haven’t conceived of.
Which of course gets me to another major point:
You need to be familiar with Unix. Not just familiar, actually comfortable using it. The overwhelming majority of all internet services, including the core internet infrastructure, all run on Unix.
(Side note:Linux != Unix)
You need to be able to operate in the Unix environment and use all the tools: sed , awk , grep , sort , uniq , and everything in between. These tools and their semantics permeate all the development tools used in software engineering, in operations, in other applications.
Here at Stevens, you can get an account on linux-lab.cs.stevens-tech.edu ; if you use a Mac, live in the terminal; install NetBSD or Linux on your laptop. Don’t use it as another platform to use occasionally; use it as your main, your only OS.
Know your editor. You can pick any one you like, but you need to know at least the basics of vi(1) . Don’t pick a trivial text editor for actual work. Know your editor. Be efficient in how to use it to write code.
Know about online references, automatic lookup of function definitions, tags files, and moving around your file without using a mouse.
Your editor is powerful. You spend the majority of your day in your editor, you should be efficient in using it. Invest the time to follow a tutorial.
When you write code, make sure you understand how to best debug your code. Most people do printf-based debugging: you write some code, you run it, it fails, you open the editor again and add a " printf('Here./n'); " statement, compile it again, run it again, and then repeat.
This is terribly inefficient. Instead, learn to use a debugger that allows you to inspect your program (either while it’s running or after it has already failed by way of a core file) without having to modify it.
Debugging code is difficult. You will spend significantly more time debugging code — yours and other people’s — than writing code. So you should be good at it! Debugging consists of the process of slowly, painstakingly, discovering what of the many things you thought were true are not. Computers are great: they do exactly what you tell them to. They do not do what you think you told them to, however. So you need to figure out what you thought was the case versus what you told the computer.
To do this efficiently, you should have a hypothesis for any problem you encounter. Do not blindly poke around and hope to find the bug: have a theory of what you think is the case, verify whether or not that is true, repeat.
Often times the first step in debugging is figuring out just where exactly in the code the bug takes place. Divide and conquer is a very effective way to find your bug. Don’t randomly look here or there, but make specific, educated guesses that eliminate specific possibilities so you don’t have to waste your time on chasing down dead ends.
Know how to trace the execution of a program.
Sometimes you need to debug the behaviour of a program you either do not have the code for, or where diving into the code base is just not likely to be efficient. You should be able to trace the execution of any program to track down,for example, which files are opened in which order. You may be able to use a debugger for this, too, by attaching to a running process, but you also want to be familiar with the various ‘trace’ tools ( strace(1) on Linux, dtrace(1) on Solaris/FreeBSD/OS X, ktrace(1) on NetBSD, …). They are your friends.
In school and in your personal projects, it’s fun and exciting to try out new languages, new frameworks, new libraries. When you want to build a reliable product or infrastructure, however, "exciting" is very much the opposite of what you want. Use the most boring technology . Use what you already know by heart. More importantly:
Don’t try to be clever. Clever code is complex, often complicated. Clever code is hard to debug, and often has unexpected consequences. Even if you are really proud about having written code that looks like:
print foo.join(map(lambda x: f(x), g(y)))
This becomes hard to debug or understand just a few minutes after you’ve closed the file.
Good code is not clever. Good code does not make the reader go "wait, what? Huh. Oh, hmm, maybe. I guess that’s cool. But what if…".
Good code is easy to understand, simple. Simplify! Murder your darlings.
There’s some really good advice in here. It applies to all programming, not just python. Make sure you understand the reasons behind them.
Write code that is easy to read. Readability is important because other people will have to read your code. That includes your own future self. A few days after you’ve written your code, you will no longer fully remember why you wrote what you wrote.
You can increase readability using a few simple rules, some of which you’ve probably heard before. Building software is tricky, and it requires you to be able to keep in your mind a full mental model of whatever functionality you are currently implementing. You should break your code into distinct chunks that are self-contained and not too complex.
A good rule of thumb is that if your code does not fit into a few screenfulls of your editor, it’s too complex and should be refactored.
By the way, that is why it’s useful to use a standardized terminal window, instead of a giant, vertical full screen window on a 30inch monitor.
Another sign that it’s time to refactor is when your code falls off the side of the editor window because you are indented so far to the right that you can’t get a function name in.
And renaming your functions and variables to single-characters does not count. That’s a terrible idea.
Your code should be easy to read and descriptive . You are not charged per character, so don’t use single-character variables or variables where you removed the vowels. Why do programmers hate vowels so much?
Name your functions after the actions they perform and the return type they have. Be consistent in naming conventions for your variables.
Once you begin giving your functions descriptive names, you also more easily discover when you are stuffing too much functionality into a single function when it should instead be multiple smaller functions. Whenever your function name contains the word "and", you probably want to break it into two separate functions.
This is tough for CS students. In most classes they are told that they should comment all their code all the time. But that’s not actually a good thing, because it leads to things like:
Sure, it’s superfluous, but why is this bad? Well, for one thing, comments and code are two different languages that our brain needs to switch back and forth between when we read code. This requires context switches and is difficult. Building a mental model of what a given piece of code does is difficult if you get disrupted by unnecessary and distracting comments.
But a more important aspect is that code and comments all too easily diverge over time. Suppose you encounter a bug in the code and realize it should actually be incremented by 4, not 2. You know what happens next?
People forget to update comments all the time. Now suppose you later come across this piece of code. You’d be inclined to believe that there’s a bug here, since clearly x should be incremented by 2, not by 4. And so you go and change it back.
Comments are useful when they describe why the code is necessary, not what it does . If you write your code in a descriptive manner with clear function and variable names, then it’ll be easy for anybody who is familiar with the problem space and the programming language to understand what it does.
So write your code such that it is self-explanatory and does not require any comments, except to explain the why .
Writing comments where useful and clear, readable code is a major help to those debugging your code later. The more you help them understand what you’re doing and why, the better. Writing code is only one aspect of this.
All your programs should come with adequate documentation. This should include a description of what the program does and how to use it. Write an actual manual page — it’s not very difficult, but it helps you define your program.
(I actually start out by writing the manual page for any of the tools I write, because that helps me figure out how I anticipate the tool to be used, what the user interface is like, what users want to do with the tool etc.)
Manual pages are written in one of the ‘roff’ dialects. This is just another markup language. The easiest way to write a manual page is to copy an existing one and change it.
I mentioned that you should simplify your code such that each program, tool, function does one thing and one thing only. This also allows you to test your code in an automated fashion. For every bit of new functionality, you begin by thinking about how you can test the code to verify it actually does what you think it should do. Then you write the test. Only after you have a test do you begin writing code that performs the functionality.
By following this method , you are accumulating an increasingly complete test suite, which allows you to verify that after you have added new functionality that you didn’t break anything else. This can then later be included in more sophisticated frameworks to assert correctness of code in larger projects.
There is no need to write "Success." after your tool did what it was asked to do. Similarly, there is no good use in saying "Error!" when you encountered a problem. Just saying "Error!" does not help the user.
Instead, you should make sure to provide precise, meaningful error messages. Most programming languages provide you with library functions to do that for you (see perror(3)/strerror(3) ). Be descriptive in what went wrong and what the cause of the error was.
Having your programs generate meaningful error messages helps in debugging them. It helps others debug them, which then allows them to write a bug report.
Bug reports are great, when you’re a developer, because they tell you about a problem in your software that you can fix for your users. Unfortunately, it can be really difficult to reproduce a given problem. Developers usually spend more time trying to reproduce the problem than they are in fixing the actual bug.
A good bug report includes all the relevant information (such as software version, OS, libraries), what the user tried to do, what they expected to happen, what actually happened, and what the exact error was.
When generating an error report, do not handwavingly say "and then my internet broke", or "then it said something about not being able to read some file somewhere", but be specific. Copy the exact error from the command. Tell the developers what you tried to do to fix the problem.
The best part about writing a thorough and accurate bug report is that you most likely find that you did something wrong, or you find out what the problem was. You end up better understanding what you’re doing, which is, generally speaking, a good thing.
When you write code, it’s all too easy to search on the internet for a solution, to find some code on StackOverflow or elsewhere and to just copy and paste it. When you do that, make sure that you actually understand what the code in question does, why it works and why it was written the way it was written.
Don’t fix bugs and ship a new version without actually understanding why the bug fix works, what was wrong before, and what possible edge conditions the new code may or may not account for.
Remember also that most of the answers given on StackOverflow or the Internet in general are likely given by people like you. Your impostor syndrome makes you think everybody else knows what they’re talking about; their Dunning-Kruger makes them think they’re experts. So be careful about what you copy and paste.
Now lazy cheating — such as copying things from StackOverflow — is terrible, and will get your professor upset and annoyed. (And upset and/or annoyed professors do not give good grades.) So if you plan on cheating, you better do it well. Just googling and handing in somebody else’s code is dumb; you don’t learn anything, and if you’re able to find the code on the internet to hand in, so will your professor.
Now if, on the other hand, you find the code you want, and you rewrite it such that it fits in nicely with your framework, uses your coding style, naming convention, includes your own comments etc., then you’ll likely have actually (accidentally) learned something.
Contrary to the common stereotype, software development is an intensely social activity. You very rarely are hacking away all by yourself to produce a final product. You almost always work in a team, together with several other individuals and interacting with dozens or hundreds of people throughout the development cycle.
This requires you to be able to communicate efficiently with all your colleagues. One method of communicating is via code — I covered that above, with regards to producing clear, readable code; by providing good documentation; by writing high quality bug reports.
But you need to discuss your code more immediately with your peers. You need to be able to write pull requests, or to provide code reviews and design feedback. You need to understand the expectations of your users and communicate with them about your release cycle and bug fixes.
A lot of this happens not in synchronous face-to-face encounters, which may make it more difficult for some people to communicate effectively. You need to learn how to:
- communicate in synchronous online chat (IRC, Hipchat, Slack, …)
- communicate via video conferencing (Skype, Google Hangouts, …) even with poor connections
- communicate asynchronously via email
Seriously, it appears that students do not know how to write an email.
An email should:
- contain a greeting, unless you reply to a previous mail
- contain full, grammatically correct sentences using proper spelling
- beplain text; "my replies are in blue" does not work
- be well-formatted and readable
- use paragraph and line breaks
- onlyquote what is needed
You should be aware that emails that you write to one person are forwarded or quoted in mails to somebody else. Messages on your online chat platform are likely logged (and may be read back in a future court case ). Websites and comments on forums are archived and available even after the site in question has long gone belly-up.
When you are communicating with other people, it’s easy to forget the most obvious thing about other people: namely, that they are, in fact, people. You should treat them as such.
Contrary to what it may seem to you, nobody gets up in the morning and plans on making other people’s lives miserable. Nobody. Well, except for Donald Trump. Fuck that guy. What a dick.
Don’t be a dick. People generally mean well .
When you get into arguments with your users, your peers, other programmers, and people you’ve never met in real life; when you begin to exchange arguments via email, or you begin the passive-aggressive ticket closing-re-opening game , remind yourself that other people probably have their own reasons for how they act, and that those reasons may well make sense to them.
You are likely missing some information and you do not fully understand what their priorities are. Everybody else’s job is more complicated than you think.
This is especially difficult when you are dealing with mistakes — your own, and other people’s.
You should always stand by your mistakes. They’re one of the most effective (although not necessarily most efficient) ways to really learn something. Do not try to deflect blame if a mistake was yours. Take responsibility, seek to understand the decision making process that lead you to the mistake you made, and learn from it.
People will respect you more , not less, if you own your mistakes. Everybody makes them! Becoming defensive and trying to blame the library, the process, other people, or the universe at large is not going to play out well in your favor.
Plus, spectacular errors give you someinteresting stories to tell years later.
One part of owning your mistakes is that if you break something, you have to fix it, right? If you issue a pull request to add a feature or to fix one thing, but that happens to break something else, then yes, it’s your responsibility to fix that.
Now here’s an interesting corollary I’ve observed:
Every organization has heaps and heaps of stale, broken, legacy code. Code that nobody wants to touch, nobody wants to maintain, or sometimes even just code that is no longer interesting, "exciting" to the developers (see above re being boring!).
Now you come along and you find a problem. You discover a bug. You dig in, you find the cause, you fix it. "Pull requests welcome!" they said.
Well, guess what? You are now the proud new owner of the entire code base. Any future issues will be brought to you, because "you touched it last", or because you’re the only one who still understands it, or you’re the only one who cared enough to fix it.
No, the lesson here is not "don’t fix other people’s stuff". There is no real lesson here. Just something you should be aware of.
Studies have shown that 75% of the total cost of ownership of a given piece of software is not the initial development, but the ongoing maintenance . Your code isn’t finished when you release your product. That’s when it begins its life!
And you’re not done there. You need to make sure you understand how your product is deployed, configured, monitored, upgraded, etc.
Which is why it’s so important that your code is simple, readable, and has good documentation…
This is important in helping you learn all of the above: You should practice programming as much as you can, but it’s not useful to just sit and hack all by yourself. Software development is a social process, and you need to learn to collaborate. You need to be able to collaborate with different people from different backgrounds with different capabilities.
Participating in Open Source helps you learn all that. And you don’t need to be a top-notch programmer to help and have an impact: almost all open source projects need people to help with documentation, with testing, with infrastructure maintenance, website administration, …
Join a community of a product you use a lot. Join their mailing lists, hang out on their IRC channels, write bug reports, submit patches.
Open Source projects are communities. Different communities have different styles of communication, different values. Pick your communities carefully. The people you surround yourself with influence you, especially when you agree on everything!
This is perhaps more important nowadays than it was even a few years ago. Social media allows you to interact entirely with people who agree with you. And every signal you get reinforces your own opinions. All of a sudden, you only see content that confirms what you already believe, which makes it impossible for you to build diverse and balanced opinions.
This effect is called the ‘ filter bubble ‘. You should carefully be aware of this effect and actively seek out opposing opinions and articles on things that are not already within your own area of interest.
Filter bubbles exist on social media, but also in real life. People tend to get together based on their interests, but if the groups you’re joining are too homogeneous, then it’s maybe time to seek out diversity.
If you’re in a particularly homogeneous community — say, a Computer Science club at a college or university — ask yourself why it is so homogeneous. Are there fewer women because the subject matter is simply something that women don’t understand or have no interest in? Are there few people of color because even though they might be interested, they might not feel comfortable as the only outsider in a homogeneous community?
Does your open source project welcome people from all backgrounds or do they buy into the meritocracy myth while simultaneously disparaging or dismissing contributions from people who are not just like them?
Who here is a feminist? (Two, three hands go up, timidly.)
Wait, let me ask another way around: who here thinks that all people, regardless of gender, have the same rights and should be treated equally? (All hands go up.)
Congratulations, you are all feminists.
Again, open source projects can teach you valuable lessons and expose you to a lot of diversity, but it can also tighten your filter bubble. Be aware of these factors. And remember: the internet doesn’t forget. How you engage others here is strictly "on the record".
Open Source participation is an important part of your resume. Employers are looking at your contributions, your participation in the Open Source world, too! GitHub may not replace your resume (there are a number of fallacies here), but if you have code you’ve written, patches you’ve submitted or any other contribution in the open, we will look at that. It helps us a lot more than knowing that you took Data Structures and Algorithms I in your Spring semester.
Having contributed to a large open source project, being an active member of a thriving community speaks volumes about your ability to communicate with others, and it will be looked upon favorably when interviewing.
Talking about interviewing, here’s something that’s useful to know:
The skills to pass an interview are not the same skill as those needed to do the job. Nobody knows how to do it well. Most of the time it’s just a needlessly adversarial face-off where the interview panel shows you how much more they know about computering than you do.
Many people have tried to fix the interview process, but you still see meaningless brain teasers, and ad-hoc white board programming and questions about algorithmic complexity, and then you get passed over for not being a "culture fit".
(Yes, there’s another, much longer rant brewing here.)
There’s plenty of good advice out there on how to perform better in interviews, but I think one of the most important aspects is to make a good first impression. People — consciously or subconsciously — will make a hire/no-hire decision within the first ten minutes of your conversation. The remainder is a waste; if they don’t like you, they won’t get persuaded otherwise just because you wrote a reasonable implementation of Huffman compression on the whiteboard.
So be sure to make a positive, enthusiastic, friendly, polite, competent first impression. Back up your resume with the Open Source experience you have.
Ok, so let’s suppose you passed the test and get a job offer. Here comes the next wave of bullshit: salary negotiations.
The company will ask you for your salary history when making an offer. Now as a new graduate, you may not have a salary history, but they will still ask you to come up with a number. That’s bullshit. Resists, push back. Let them name a number first.
In order to understand what salary you can negotiate, you need to know what you’re worth . Glassdoor can help, to some degree, but you also should get feedback from your peers, colleagues, mentors, your open source buddies to know the average salary for somebody in your position with your experience and skills in the given job.
This is difficult — especially for new graduates — because often times people are not comfortable sharing how much they make. Especially once you’re in a job, people like you to not talk about this with your colleagues, but that just fosters the imbalance between employer and employee. Salary transparency empowers employers and leads to fairer wages, especially for minorities.
There are plenty of articles everywhere on how to better negotiate your salary, but keep in mind:
As long as you reveal your salary history in future job interviews, that is. The next company will offer you $currentSalary + 20% if you’re lucky.
On the other hand, your career path need not be written in stone:
One of the great things in this field is that you can switch careers fairly easily. You can be a frontend developer becoming a backend developer becoming a systems architect turning information security professional turning project manager turning CTO and anything in between.
But every job you pick along the way will pigeon-hole you. The better you get at one thing, the easier it will be to continue doing that one thing. This is a bit of the same effect as the filter bubble I mentioned earlier. Switching career paths is difficult, but certainly possible. The broader your interests and expertise, the easier that is for you. Seek diversity.
Don’t fool yourself into thinking you can multitask. You can’t listen to a talk, or watch a TV program and get meaningful work done.
Anything that requires concentration requires your full attention.
Sure, there are some things that don’t require your full attention, but ideally you’d try to eliminate most of those.
You want to get work done? Disable your email, log out of Facebook and Twitter. Turn off the internet altogether. (Some of my most productive hours are on a plane, when nobody bothers me, I don’t have any distractions, and there’s no internet .)
Working more hours does not mean more work gets done.
On a good day, you get about 3 to 4 hours of actual work done. If you’re lucky. The rest of the time is spent "doing email" or attending meetings. Those are distractions. Cut them down as much as you can.
Don’t write a single line of code that you don’t need. You’re likely just adding new bugs, unneeded complexity. Simplify.
The majority of security problems in your applications stem from not properly validated input. See ‘ Little Bobby Tables ‘.
Never accept or pass input from the user (including other programs) without asserting that it’s well-formatted and valid.
Input validation cannot be done by creating a list of things that you do not want to accept — you will never catch all the edge cases. You need to explicitly whitelist the patterns you want to accept, reject anything else.
All your programs need to have security added in the beginning. Include your security team in the planning phase, when you begin to figure out what you want to do.
You cannot ship a product and then later on rub some crypto on it to make it "secure". Regularly talk to your security team. (If you’re on the security team: regularly talk to your developers.)
Time is an illusion . Sometimes an hour doesn’t exist (two nights ago), sometimes an hour repeats. Sometimes a minute has 61 seconds. Timezones shift. Days disappear.
At any but the most minimal scale will you operate across multiple timezones, most likely across different countries. Event correlation and understanding what happened when becomes a mess when some events are logged in Pacific Time, some in GMT, and some in Indian time, but you are sitting in New York.
Times are represented differently in different languages. Use the W3CDTF format , use a 24 hour clock.
Know how to use revision control. Nobody knows how git works. That’s ok. But you need to know enough to be able to work with other people, with different teams, to understand the basics of branching and merging code. These concepts apply to more than just git.
(See also: Notes on Distributed Systems for Young Bloods )
Understand how IP addresses are allocated, how top-level domains are created and maintained, how peering points work, as well as the limitations imposed by physical laws.
(If you’re interested, you can join us tonight for my lecture on networking in my SysAdmin class, where we cover a fair bit of how the internet works.)
When things break down and nothing seems to make sense, the core problem almost always is hidden somewhere in the DNS. You need to really understand how this works. And please don’t monkey around with /etc/hosts — you’re just going to make your life miserable and incur hours of troubleshooting and debugging when you forget what you added there.
You can troubleshoot applications and debug code all you like — sometimes, you need to be able to observe exactly what’s going on on the network. Wireshark and other tools are nice, but you should be able to read a pcap(3) file without those.
You should understand the OSI stack and know which layer you’re operating on.
The internet is awesome. It doesn’t have a central controlling body, but there is the Internet Engineering Task Force (IETF), which oversees and steers the development of the standards and protocols.
Just like the various Open Source projects, this is an open community and you can be part of it. You can help keep the internet open and you can help make it awesomer. You can join the open mailing lists of some of the working groups and just observe how policy is made, how standards are developed, and you can then partake and help move the Internet forward.
You are privileged. You are fortunate enough to be able to afford this education, to learn how the internet works, to learn how to write software, and you are in a privileged position compared to the overwhelming majority of people.
With this privilege comes the responsibility to ensure that the data of your users is kept safe, that secure standards are developed for the betterment of society, to ensure our community acts in the public interest .
By the way, there are two more layers to the OSI stack which most schools do not (adequately) cover:
This is important to understand. No matter what you do, your work is political. There are inter-office politics, politics between industry competitors, between governments and industry, between governments and governments, and no matter how much you tell yourself that you’re just here to solve a technical problem, it will impact your work.
For this reason, you should also have an interest in and follow regional, national, and international politics. You cannot stick your head in the sand and pretend it doesn’t affect you. Especially not today.
And finally, some common sense. Wash your hands.
- Don’t do drugs.
- Use an ad-blocker.
- Stay in school.
- Disable Flash.
- Anonymity is important. (Speak up!)
- Yes means yes.
- Don’t get phished.
- Black lives matter.
- Use a password manager. (I happen to like 1Password .)
- Wear sunscreen.
Also: don’t upload your passwords or ssh keys to GitHub, please.
March 14th, 2016