37 KiB
Education of a Programmer
When I left Microsoft in October 2016 after almost 21 years there and almost 35 years in the industry, I took some time to reflect on what I had learned over all those years. This is a lightly edited version of that post. Pardon the length!
There are an amazing number of things you need to know to be a proficient programmer — details of languages, APIs, algorithms, data structures, systems and tools. These things change all the time — new languages and programming environments spring up and there always seems to be some hot new tool or language that “everyone” is using. It is important to stay current and proficient. A carpenter needs to know how to pick the right hammer and nail for the job and needs to be competent at driving the nail straight and true.
At the same time, I’ve found that there are some concepts and strategies that are applicable over a wide range of scenarios and across decades. We have seen multiple orders of magnitude change in the performance and capability of our underlying devices and yet certain ways of thinking about the design of systems still say relevant. These are more fundamental than any specific implementation. Understanding these recurring themes is hugely helpful in both the analysis and design of the complex systems we build.
Humility and Ego
This is not limited to programming, but in an area like computing which exhibits so much constant change, one needs a healthy balance of humility and ego. There is always more to learn and there is always someone who can help you learn it — if you are willing and open to that learning. One needs both the humility to recognize and acknowledge what you don’t know and the ego that gives you confidence to master a new area and apply what you already know. The biggest challenges I have seen are when someone works in a single deep area for a long time and “forgets” how good they are at learning new things. The best learning comes from actually getting hands dirty and building something, even if it is just a prototype or hack. The best programmers I know have had both a broad understanding of technology while at the same time have taken the time to go deep into some technology and become the expert. The deepest learning happens when you struggle with truly hard problems.
End to End Argument
Back in 1981, Jerry Saltzer, Dave Reed and Dave Clark were doing early work on the Internet and distributed systems and wrote up their classic description of the end to end argument. There is much misinformation out there on the Internet so it can be useful to go back and read the original paper. They were humble in not claiming invention — from their perspective this was a common engineering strategy that applies in many areas, not just in communications. They were simply writing it down and gathering examples. A minor paraphrasing is:
When implementing some function in a system, it can be implemented correctly and completely only with the knowledge and participation of the endpoints of the system. In some cases, a partial implementation in some internal component of the system may be important for performance reasons.
The SRC paper calls this an “argument”, although it has been elevated to a “principle” on Wikipedia and in other places. In fact, it is better to think of it as an argument — as they detail, one of the hardest problem for a system designer is to determine how to divide responsibilities between components of a system. This ends up being a discussion that involves weighing the pros and cons as you divide up functionality, isolate complexity and try to design a reliable, performant system that will be flexible to evolving requirements. There is no simple set of rules to follow.
Much of the discussion on the Internet focuses on communications systems, but the end-to-end argument applies in a much wider set of circumstances. One example in distributed systems is the idea of “eventual consistency”. An eventually consistent system can optimize and simplify by letting elements of the system get into a temporarily inconsistent state, knowing that there is a larger end-to-end process that can resolve these inconsistencies. I like the example of a scaled-out ordering system (e.g. as used by Amazon) that doesn’t require every request go through a central inventory control choke point. This lack of a central control point might allow two endpoints to sell the same last book copy, but the overall system needs some type of resolution system in any case, e.g. by notifying the customer that the book has been backordered. That last book might end up getting run over by a forklift in the warehouse before the order is fulfilled anyway. Once you realize an end-to-end resolution system is required and is in place, the internal design of the system can be optimized to take advantage of it.
In fact, it is this design flexibility in the service of either ongoing performance optimization or delivering other system features that makes this end-to-end approach so powerful. End-to-end thinking often allows internal performance flexibility which makes the overall system more robust and adaptable to changes in the characteristics of each of the components. This makes an end-to-end approach “anti-fragile” and resilient to change over time.
An implication of the end-to-end approach is that you want to be extremely careful about adding layers and functionality that eliminates overall performance flexibility. (Or other flexibility, but performance, especially latency, tends to be special.) If you expose the raw performance of the layers you are built on, end-to-end approaches can take advantage of that performance to optimize for their specific requirements. If you chew up that performance, even in the service of providing significant value-add functionality, you eliminate design flexibility.
The end-to-end argument intersects with organizational design when you have a system that is large and complex enough to assign whole teams to internal components. The natural tendency of those teams is to extend the functionality of those components, often in ways that start to eliminate design flexibility for applications trying to deliver end-to-end functionality built on top of them.
One of the challenges in applying the end-to-end approach is determining where the end is. “Little fleas have lesser fleas… and so on ad infinitum.”
Concentrating Complexity
Coding is an incredibly precise art, with each line of execution required for correct operation of the program. But this is misleading. Programs are not uniform in the overall complexity of their components or the complexity of how those components interact. The most robust programs isolate complexity in a way that lets significant parts of the system appear simple and straightforward and interact in simple ways with other components in the system. Complexity hiding can be isomorphic with other design approaches like information hiding and data abstraction but I find there is a different design sensibility if you really focus on identifying where the complexity lies and how you are isolating it.
The example I’ve returned to over and over again in my writing is the screen repaint algorithm that was used by early character video terminal editors like VI and EMACS. The early video terminals implemented control sequences for the core action of painting characters as well as additional display functions to optimize redisplay like scrolling the current lines up or down or inserting new lines or moving characters within a line. Each of those commands had different costs and those costs varied across different manufacturer’s devices. (See TERMCAP for links to code and a fuller history.) A full-screen application like a text editor wanted to update the screen as quickly as possible and therefore needed to optimize its use of these control sequences to transition the screen from one state to another.
These applications were designed so this underlying complexity was hidden. The parts of the system that modify the text buffer (where most innovation in functionality happens) completely ignore how these changes are converted into screen update commands. This is possible because the performance cost of computing the optimal set of updates for any change in the content is swamped by the performance cost of actually executing the update commands on the terminal itself. It is a common pattern in systems design that performance analysis plays a key part in determining how and where to hide complexity. The screen update process can be asynchronous to the changes in the underlying text buffer and can be independent of the actual historical sequence of changes to the buffer. It is not important how the buffer changed, but only what changed. This combination of asynchronous coupling, elimination of the combinatorics of historical path dependence in the interaction between components and having a natural way for interactions to efficiently batch together are common characteristics used to hide coupling complexity.
Success in hiding complexity is determined not by the component doing the hiding but by the consumers of that component. This is one reason why it is often so critical for a component provider to actually be responsible for at least some piece of the end-to-end use of that component. They need to have clear optics into how the rest of the system interacts with their component and how (and whether) complexity leaks out. This often shows up as feedback like “this component is hard to use” — which typically means that it is not effectively hiding the internal complexity or did not pick a functional boundary that was amenable to hiding that complexity.
Layering and Componentization
It is the fundamental role of a system designer to determine how to break down a system into components and layers; to make decisions about what to build and what to pick up from elsewhere. Open Source may keep money from changing hands in this “build vs. buy” decision but the dynamics are the same. An important element in large scale engineering is understanding how these decisions will play out over time. Change fundamentally underlies everything we do as programmers, so these design choices are not only evaluated in the moment, but are evaluated in the years to come as the product continues to evolve.
Here are a few things about system decomposition that end up having a large element of time in them and therefore tend to take longer to learn and appreciate.
- Layers are leaky. Layers (or abstractions) are fundamentally leaky. These leaks have consequences immediately but also have consequences over time, in two ways. One consequence is that the characteristics of the layer leak through and permeate more of the system than you realize. These might be assumptions about specific performance characteristics or behavior ordering that is not an explicit part of the layer contract. This means that you generally are more vulnerable to changes in the internal behavior of the component that you understood. A second consequence is it also means you are more dependent on that internal behavior than is obvious, so if you consider changing that layer the consequences and challenges are probably larger than you thought.
- Layers are too functional. It is almost a truism that a component you adopt will have more functionality than you actually require. In some cases, the decision to use it is based on leveraging that functionality for future uses. You adopt specifically because you want to “get on the train” and leverage the ongoing work that will go into that component. There are a few consequences of building on this highly functional layer. 1) The component will often make trade-offs that are biased by functionality that you do not actually require. 2) The component will embed complexity and constraints because of functionality you do not require and those constraints will impede future evolution of that component. 3) There will be more surface area to leak into your application. Some of that leakage will be due to true “leaky abstractions” and some will be explicit (but generally poorly controlled) increased dependence on the full capabilities of the component. Office is big enough that we found that for any layer we built on, we eventually fully explored its functionality in some part of the system. While that might appear to be positive (we are more completely leveraging the component), all uses are not equally valuable. So we end up having a massive cost to move from one layer to another based on this long-tail of often lower value and poorly recognized use cases. 4) The additional functionality creates complexity and opportunities for misuse. An XML validation API we used would optionally dynamically download the schema definition if it was specified as part of the XML tree. This was mistakenly turned on in our basic file parsing code which resulted in both a massive performance degradation as well as an (unintentional) distributed denial of service attack on a w3c.org web server. (These are colloquially known as “land mine” APIs.)
- Layers get replaced. Requirements evolve, systems evolve, components are abandoned. You eventually need to replace that layer or component. This is true for external component dependencies as well as internal ones. This means that the issues above will end up becoming important.
- Your build vs. buy decision will change. This is partly a corollary of above. This does not mean the decision to build or buy was wrong at the time. Often there was no appropriate component when you started and it only becomes available later. Or alternatively, you use a component but eventually find that it does not match your evolving requirements and your requirements are narrow enough, well-understood or so core to your value proposition that it makes sense to own it yourself. It does mean that you need to be just as concerned about leaky layers permeating more of the system for layers you build as well as for layers you adopt.
- Layers get thick. As soon as you have defined a layer, it starts to accrete functionality. The layer is the natural throttle point to optimize for your usage patterns. The difficulty with a thick layer is that it tends to reduce your ability to leverage ongoing innovation in underlying layers. In some sense this is why OS companies hate thick layers built on top of their core evolving functionality — the pace at which innovation can be adopted is inherently slowed. One disciplined approach to avoid this is to disallow any additional state storage in an adaptor layer. Microsoft Foundation Classes took this general approach in building on top of Win32. It is inevitably cheaper in the short term to just accrete functionality on to an existing layer (leading to all the eventual problems above) rather than refactoring and recomponentizing. A system designer who understands this looks for opportunities to break apart and simplify components rather than accrete more and more functionality within them.
Einsteinian Universe
I had been designing asynchronous distributed systems for decades but was struck by this quote from Pat Helland, a SQL architect, at an internal Microsoft talk. “We live in an Einsteinian universe — there is no such thing as simultaneity. “ When building distributed systems — and virtually everything we build is a distributed system — you cannot hide the distributed nature of the system. It’s just physics. This is one of the reasons I’ve always felt Remote Procedure Call, and especially “transparent” RPC that explicitly tries to hide the distributed nature of the interaction, is fundamentally wrong-headed. You need to embrace the distributed nature of the system since the implications almost always need to be plumbed completely through the system design and into the user experience.
Embracing the distributed nature of the system leads to a number of things:
- You think through the implications to the user experience from the start rather than trying to patch on error handling, cancellation and status reporting as an afterthought.
- You use asynchronous techniques to couple components. Synchronous coupling is impossible. If something appears synchronous, it’s because some internal layer has tried to hide the asynchrony and in doing so has obscured (but definitely not hidden) a fundamental characteristic of the runtime behavior of the system.
- You recognize and explicitly design for interacting state machines and that these states represent robust long-lived internal system states (rather than ad-hoc, ephemeral and undiscoverable state encoded by the value of variables in a deep call stack).
- You recognize that failure is expected. The only guaranteed way to detect failure in a distributed system is to simply decide you have waited “too long”. This naturally means that cancellation is first-class. Some layer of the system (perhaps plumbed through to the user) will need to decide it has waited too long and cancel the interaction. Cancelling is only about reestablishing local state and reclaiming local resources — there is no way to reliably propagate that cancellation through the system. It can sometimes be useful to have a low-cost, unreliable way to attempt to propagate cancellation as a performance optimization.
- You recognize that cancellation is not rollback since it is just reclaiming local resources and state. If rollback is necessary, it needs to be an end-to-end feature.
- You accept that you can never really know the state of a distributed component. As soon as you discover the state, it may have changed. When you send an operation, it may be lost in transit, it might be processed but the response is lost, or it may take some significant amount of time to process so the remote state ultimately transitions at some arbitrary time in the future. This leads to approaches like idempotent operations and the ability to robustly and efficiently rediscover remote state rather than expecting that distributed components can reliably track state in parallel. The concept of “eventual consistency” succinctly captures many of these ideas.
I like to say you should “revel in the asynchrony”. Rather than trying to hide it, you accept it and design for it. When you see a technique like idempotency or immutability, you recognize them as ways of embracing the fundamental nature of the universe, not just one more design tool in your toolbox.
Performance
I am sure Don Knuth is horrified by how misunderstood his partial quote “Premature optimization is the root of all evil” has been. In fact, performance, and the incredible exponential improvements in performance that have continued for over 6 decades (or more than 10 decades depending on how willing you are to project these trends through discrete transistors, vacuum tubes and electromechanical relays), underlie all of the amazing innovation we have seen in our industry and all the change rippling through the economy as “software eats the world”.
A key thing to recognize about this exponential change is that while all components of the system are experiencing exponential change, these exponentials are divergent. So the rate of increase in capacity of a hard disk changes at a different rate from the capacity of memory or the speed of the CPU or the latency between memory and CPU. Even when trends are driven by the same underlying technology, exponentials diverge. Latency improvements fundamentally trail bandwidth improvements. Exponential change tends to look linear when you are close to it or over short periods but the effects over time can be overwhelming. This overwhelming change in the relationship between the performance of components of the system forces reevaluation of design decisions on a regular basis.
A consequence of this is that design decisions that made sense at one point no longer make sense after a few years. Or in some cases an approach that made sense two decades ago starts to look like a good trade-off again. Modern memory mapping has characteristics that look more like process swapping of the early time-sharing days than it does like demand paging. (This does sometimes result in old codgers like myself claiming that “that’s just the same approach we used back in ‘75” — ignoring the fact that it didn’t make sense for 40 years and now does again because some balance between two components — maybe flash and NAND rather than disk and core memory — has come to resemble a previous relationship).
Important transitions happen when these exponentials cross human constraints. So you move from a limit of two to the sixteenth characters (which a single user can type in a few hours) to two to the thirty-second (which is beyond what a single person can type). So you can capture a digital image with higher resolution than the human eye can perceive. Or you can store an entire music collection on a hard disk small enough to fit in your pocket. Or you can store a digitized video recording on a hard disk. And then later the ability to stream that recording in real time makes it possible to “record” it by storing it once centrally rather than repeatedly on thousands of local hard disks.
The things that stay as a fundamental constraint are three dimensions and the speed of light. We’re back to that Einsteinian universe. We will always have memory hierarchies — they are fundamental to the laws of physics. You will always have stable storage and IO, memory, computation and communications. The relative capacity, latency and bandwidth of these elements will change, but the system is always about how these elements fit together and the balance and tradeoffs between them. Jim Gray was the master of this analysis.
Another consequence of the fundamentals of 3D and the speed of light is that much of performance analysis is about three things: locality, locality, locality. Whether it is packing data on disk, managing processor cache hierarchies, or coalescing data into a communications packet, how data is packed together, the patterns for how you touch that data with locality over time and the patterns of how you transfer that data between components is fundamental to performance. Focusing on less code operating on less data with more locality over space and time is a good way to cut through the noise.
Jon Devaan used to say “design the data, not the code”. This also generally means when looking at the structure of a system, I’m less interested in seeing how the code interacts — I want to see how the data interacts and flows. If someone tries to explain a system by describing the code structure and does not understand the rate and volume of data flow, they do not understand the system.
A memory hierarchy also implies we will always have caches — even if some system layer is trying to hide it. Caches are fundamental but also dangerous. Caches are trying to leverage the runtime behavior of the code to change the pattern of interaction between different components in the system. They inherently need to model that behavior, even if that model is implicit in how they fill and invalidate the cache and test for a cache hit. If the model is poor or becomes poor as the behavior changes, the cache will not operate as expected. A simple guideline is that caches must be instrumented — their behavior will degrade over time because of changing behavior of the application and the changing nature and balance of the performance characteristics of the components you are modeling. Every long-time programmer has cache horror stories.
I was lucky that my early career was spent at BBN, one of the birthplaces of the Internet. It was very natural to think about communications between asynchronous components as the natural way systems connect. Flow control and queueing theory are fundamental to communications systems and more generally the way that any asynchronous system operates. Flow control is inherently resource management (managing the capacity of a channel) but resource management is the more fundamental concern. Flow control also is inherently an end-to-end responsibility, so thinking about asynchronous systems in an end-to-end way comes very naturally. The story of buffer bloatis well worth understanding in this context because it demonstrates how lack of understanding the dynamics of end-to-end behavior coupled with technology “improvements” (larger buffers in routers) resulted in very long-running problems in the overall network infrastructure.
The concept of “light speed” is one that I’ve found useful in analyzing any system. A light speed analysis doesn’t start with the current performance, it asks “what is the best theoretical performance I could achieve with this design?” What is the real information content being transferred and at what rate of change? What is the underlying latency and bandwidth between components? A light speed analysis forces a designer to have a deeper appreciation for whether their approach could ever achieve the performance goals or whether they need to rethink their basic approach. It also forces a deeper understanding of where performance is being consumed and whether this is inherent or potentially due to some misbehavior. From a constructive point of view, it forces a system designer to understand what are the true performance characteristics of their building blocks rather than focusing on the other functional characteristics.
I spent much of my career building graphical applications. A user sitting at one end of the system defines a key constant and constraint in any such system. The human visual and nervous system is not experiencing exponential change. The system is inherently constrained, which means a system designer can leverage ( must leverage) those constraints, e.g. by virtualization (limiting how much of the underlying data model needs to be mapped into view data structures) or by limiting the rate of screen update to the perception limits of the human visual system.
The Nature of Complexity
I have struggled with complexity my entire career. Why do systems and apps get complex? Why doesn’t development within an application domain get easier over time as the infrastructure gets more powerful rather than getting harder and more constrained? In fact, one of our key approaches for managing complexity is to “walk away” and start fresh. Often new tools or languages force us to start from scratch which means that developers end up conflating the benefits of the tool with the benefits of the clean start. The clean start is what is fundamental. This is not to say that some new tool, platform or language might not be a great thing, but I can guarantee it will not solve the problem of complexity growth. The simplest way of controlling complexity growth is to build a smaller system with fewer developers.
Of course, in many cases “walking away” is not an alternative — the Office business is built on hugely valuable and complex assets. With OneNote, Office “walked away” from the complexity of Word in order to innovate along a different dimension. Sway is another example where Office decided that we needed to free ourselves from constraints in order to really leverage key environmental changes and the opportunity to take fundamentally different design approaches. With the Word, Excel and PowerPoint web apps, we decided that the linkage with our immensely valuable data formats was too fundamental to walk away from and that has served as a significant and ongoing constraint on development.
I was influenced by Fred Brook’s “No Silver Bullet” essay about accident and essence in software development. There is much irreducible complexity embedded in the essence of what the software is trying to model. I just recently re-read that essay and found it surprising on re-reading that two of the trends he imbued with the most power to impact future developer productivity were increasing emphasis on “buy” in the “build vs. buy” decision — foreshadowing the change that open-source and cloud infrastructure has had. The other trend was the move to more “organic” or “biological” incremental approaches over more purely constructivist approaches. A modern reader sees that as the shift to agile and continuous development processes. This in 1986!
I have been much taken with the work of Stuart Kauffman on the fundamental nature of complexity. Kauffman builds up from a simple model of Boolean networks (“NK models”) and then explores the application of this fundamentally mathematical construct to things like systems of interacting molecules, genetic networks, ecosystems, economic systems and (in a limited way) computer systems to understand the mathematical underpinning to emergent ordered behavior and its relationship to chaotic behavior. In a highly connected system, you inherently have a system of conflicting constraints that makes it (mathematically) hard to evolve that system forward (viewed as an optimization problem over a rugged landscape). A fundamental way of controlling this complexity is to batch the system into independent elements and limit the interconnections between elements (essentially reducing both “N” and “K” in the NK model). Of course this feels natural to a system designer applying techniques of complexity hiding, information hiding and data abstraction and using loose asynchronous coupling to limit interactions between components.
A challenge we always face is that many of the ways we want to evolve our systems cut across all dimensions. Real-time co-authoring has been a very concrete (and complex) recent example for the Office apps.
Complexity in our data models often equates with “power”. An inherent challenge in designing user experiences is that we need to map a limited set of gestures into a transition in the underlying data model state space. Increasing the dimensions of the state space inevitably creates ambiguity in the user gesture. This is “just math” which means that often times the most fundamental way to ensure that a system stays “easy to use” is to constrain the underlying data model.
Management
I started taking leadership roles in high school (student council president!) and always found it natural to take on larger responsibilities. At the same time, I was always proud that I continued to be a full-time programmer through every management stage. VP of development for Office finally pushed me over the edge and away from day-to-day programming. I’ve enjoyed returning to programming as I stepped away from that job over the last year — it is an incredibly creative and fulfilling activity (and maybe a little frustrating at times as you chase down that “last” bug).
Despite having been a “manager” for over a decade by the time I arrived at Microsoft, I really learned about management after my arrival in 1996. Microsoft reinforced that “engineering leadership is technical leadership”. This aligned with my perspective and helped me both accept and grow into larger management responsibilities.
The thing that most resonated with me on my arrival was the fundamental culture of transparency in Office. The manager’s job was to design and use transparent processes to drive the project. Transparency is not simple, automatic, or a matter of good intentions — it needs to be designed into the system. The best transparency comes by being able to track progress as the granular output of individual engineers in their day-to-day activity (work items completed, bugs opened and fixed, scenarios complete). Beware subjective red/green/yellow, thumbs-up/thumbs-down dashboards!
I used to say my job was to design feedback loops. Transparent processes provide a way for every participant in the process — from individual engineer to manager to exec to use the data being tracked to drive the process and result and understand the role they are playing in the overall project goals. Ultimately transparency ends up being a great tool for empowerment — the manager can invest more and more local control in those closest to the problem because of confidence they have visibility to the progress being made. Coordination emerges naturally.
Key to this is that the goal has actually been properly framed (including key resource constraints like ship schedule). Decision-making that needs to constantly flow up and down the management chain usually reflects poor framing of goals and constraints by management.
I was at Beyond Software when I really internalized the importance of having a singular leader over a project. The engineering manager departed (later to hire me away for FrontPage) and all four of the leads were hesitant to step into the role — not least because we did not know how long we were going to stick around. We were all very technically sharp and got along well so we decided to work as peers to lead the project. It was a mess. The one obvious problem is that we had no strategy for allocating resources between the pre-existing groups — one of the top responsibilities of management! The deep accountability one feels when you know you are personally in charge was missing. We had no leader really accountable for unifying goals and defining constraints.
I have a visceral memory of the first time I fully appreciated the importance of listening for a leader. I had just taken on the role of Group Development Manager for Word, OneNote, Publisher and Text Services. There was a significant controversy about how we were organizing the text services team and I went around to each of the key participants, heard what they had to say and then integrated and wrote up all I had heard. When I showed the write-up to one of the key participants, his reaction was “wow, you really heard what I had to say”! All of the largest issues I drove as a manager (e.g. cross-platform and the shift to continuous engineering) involved carefully listening to all the players. Listening is an active process that involves trying to understand the perspectives and then writing up what I learned and testing it to validate my understanding. When a key hard decision needed to happen, by the time the call was made everyone knew they had been heard and understood (whether they agreed with the decision or not).
It was the previous job, as FrontPage development manager, where I internalized the “operational dilemma” inherent in decision making with partial information. The longer you wait, the more information you will have to make a decision. But the longer you wait, the less flexibility you will have to actually implement it. At some point you just need to make a call.
Designing an organization involves a similar tension. You want to increase the resource domain so that a consistent prioritization framework can be applied across a larger set of resources. But the larger the resource domain, the harder it is to actually have all the information you need to make good decisions. An organizational design is about balancing these two factors. Software complicates this because characteristics of the software can cut across the design in an arbitrary dimensionality. Office has used shared teams to address both these issues (prioritization and resources) by having cross-cutting teams that can share work (add resources) with the teams they are building for.
One dirty little secret you learn as you move up the management ladder is that you and your new peers aren’t suddenly smarter because you now have more responsibility. This reinforces that the organization as a whole better be smarter than the leader at the top. Empowering every level to own their decisions within a consistent framing is the key approach to making this true. Listening and making yourself accountable to the organization for articulating and explaining the reasoning behind your decisions is another key strategy. Surprisingly, fear of making a dumb decision can be a useful motivator for ensuring you articulate your reasoning clearly and make sure you listen to all inputs.
Conclusion
At the end of my interview round for my first job out of college, the recruiter asked if I was more interested in working on “systems” or “apps”. I didn’t really understand the question. Hard, interesting problems arise at every level of the software stack and I’ve had fun plumbing all of them. Keep learning.
via: https://hackernoon.com/education-of-a-programmer-aaecf2d35312
作者： Terry Crowley 译者：译者ID 校对：校对者ID