Re-Intermediation -- Controlling the Internet as a Channel of Distribution

I saw an article on Slate about how Google wants to build the famous talking computer from Star Trek. Google doesn't want to return links that might contain the answer to your question, but rather to provide a direct answer. It's a romantic vision. I bet it motivates their engineering team. But it can never be.

There is a big, unbridgeable difference between Google and the Star Trek computer. Google wants to sell you something. If Google gets to the point where it can reasonably answer questions like, “What computer is best for me?”, or “Who has good prices on HDTV sets?”, or "What restaurants are nearby?", I won't be able to trust the answers, because they are shamelessly influenced by advertizer dollars.

What concerns me is whether I will have any choice in the matter.

The web was once touted as a powerful force for consumers, disintermediating old industries like TV networks, record studios, newspapers, and retail stores. But it is far more apt to think of the web as merely a new channel of distribution, disrupting older channels because of the internet's lower cost structure, and inserting new and voracious intermediaries between producers and consumers. Rather than share cost reduction with consumers, these new intermediaries want to capture all the savings as profit for themselves.

But an intermediary can only capture these savings if they dominate this new channel. Unfortunately, the internet makes that easy by reducing a company's brand name to a few keystrokes. Get that brand embedded in peoples' heads, and you own the internet as a channel. While Xerox fought for years to keep its name from becoming a verb, Google can laugh all the way to the bank. It may technically lose the ability to prevent a competitor naming itself Google, but it owns the google.com domain name, so who cares?

If Google can successfully provide direct answers instead of links, they will become the search engine. This gives them a huge advantage over other search engines, and enormous power over vendors of any product you may want to search for. Google will own the only path to find products. Amazon is set up as a marketplace, and is in one sense a competitor to Google, but people use Google first, before they even form the thought of buying a product.

Google is already dismantling the fence between their ad-based search results and organic results. If Google can answer conversationally, they will no doubt completely eliminate any distinction.

If Google becomes "the" marketplace, then they will also wield enormous power over vendors. They don't have to provide a direct way to sort products by price. They can sell placement in their search results, and extract a taste of every sale

Circular Buffer Container Class

C++ comes with a set of templatized container classes so developers don't have to re-implement a doubly linked list or balanced binary tree class for every project. This is a tremendous time-saver you may not appreciate unless you've had to live without. It can also be a trap that limits performance.

The roster of container classes in the standard library is thin; there are three kinds of sequence container (vector/array/string, deque, list/forward_list), and two kinds of associative container (map/multimap/set/multiset, unordered_map/unordered_multimap/unordered_set/unordered_multiset). With such limited choice, a developer who needs a container with specific behavior may be forced to accept a lot of unwanted implementation baggage.

By giving up the ability to access the sequence as a C-style array, it is possible to obtain a container that is about as fast as std::vector, but supports constant-time insertion and removal from both ends. One variation of this container is the circular buffer. Work to standardize the circular buffer is underway, and in the meantime boost has a version. The circular buffer makes a great queue data structure.

I ran boost::circular_buffer through the performance tests for sequence containers from my book Optimized C++. Here is a link to a long article reporting the results.

Only one new kind of container (the unordered associative container) has been added to C++ since 1995 when the Standard Template Library was introduced. There is a lot of interest from communities with performance issues in adding additional container classes to future versions of C++.

C++ Optimization Bibliography

I wrote a wonderful book on performance tuning in C++. Plug, plug, plug. Buy it now!

The web is full of free optimization advice. The best of these references support their findings with timing benchmarks. Intermediate in quality are references that speak to the behavior of specific processors, but don't contain timing data to allow the reader to reason about the size of the performance speedup. Least interesting are simple lists of recommendations presented without evidence. These are a few frequently-cited references I found useful.

Most people think of code tweaks when they write about optimization. The performance improvement from statement-level optimization is unlikely to make much difference unless it is magnified by being frequently executed, in a loop or frequently called function, so these tweaks are of uncertain value.

Free Resources on the Web 

Here is a free taste of the optimization drug, to whet your appetite and make you want to read my book, or to get you over those two days of the shakes while Amazon delivers your paper copy.

Short Notes

Tips for Optimizing C/C++ Code (PDF)

This short note contains two dozen short thought-mantras about optimizing code. Explanations are necessarily brief. It can serve as an, "I didn't know that!" kind of checklist.

Optimizing C and C++ Code (HTML)

Two dozen short thought-mantras on code optimization from game developers.

Optimization in C and C++: good practices, pitfalls, by Sébastien Binet (PDF)

Optimization advice on PowerPoint slides. Covers a lot of ground, focuses on linked structures, great graphics.

Programming Optimization (HTML), by Paul Hsieh

An interesting mix of general advice, plus brief specific low-level advice for optimizing statements and expressions.

Performance Optimization (HTML)

This is chapter 25 of the book Object Oriented System Design. It is notable for providing language-independent descriptions of design-time optimization techniques. It uses a dense and somewhat obsolete jargon defined elsewhere in the book, that can be a little hard to read. 

Three Optimization Tips for C++, Andrei Alexandrescu (HTML)

Anything Alexandrescu says is worth reading. This is more of a case study than a general advice piece, and contains many more than three tips. It features reduction in strength, reduction in array writing, and memoization.

This link has commentary and more measurements on the counting digits example Alexandrescu uses in his article.

Efficient C/C++ Coding Techniques, by Darren G. Moss (PDF)

This is a performance analysis of how several compilers available in 2001 generate code for common flow-of-control constructs, loops that index tables, and methods of structure composition. The findings are unsurprising; that switch is often better than if-elseif-else, that pointers are efficient in loops, and that a class with virtual multiple inheritance is expensive (if that wasn't obvious).

Writing Efficient C and C Code Optimization, Koushik Ghosh (HTML)

A grab-bag of statement- and expression-level optimization techniques, provided without supporting evidence for its effectiveness. Compilers may do some of these optimizations.

Techniques for Scientific C++, by Todd Veldhuizen (PDF)

A somewhat older, but interesting article with a grab-bag of optimization advice. Covers optimizing compile time, eliminating virtual functions, how aliasing interferes with optimization, expression templates,

C++ Optimization Strategies and Techniques (HTML) by Pete Isensee

Isensee really knows his stuff. This web page represents the state of the art in optimization circa 1999 (as he notes himself). It is slightly dated now, but still very relevant. The advice is well supported with timing measurements and examples.

Faster C++ (HTML)

Another grab-bag of statement-level optimizations, plus discussion of profiling, compiler flags, and cache issues. Good bibliography at the end.

FastC++: Coding Cpp Efficiently blog (HTML)

I haven't explored this blog extensively, but he seems to know a lot about getting the C++ compiler to use SSE extensions.

The Top 10 Myths of Video Game Optimization, Eric Preisz (HTML)

Interesting viewpoint on several well-loved optimizations. Of particular interest to me, that LUT-based algorithms may be a lose if they have reduced cache performance, since a cache miss is slower than the slowest instruction. The moral would be that small LUTs are good, but big ones can be problematical.

Mentor low Latency C or C++ code (HTML)

This is an article on verifying that programs meet their latency goals, not on optimizing C++ per se, But it has a bunch of good rules about memory allocation, and a couple of other optimization topics, covered without supporting evidence.

Performance of a Circular Buffer vs. Vector, Deque, and List (HTML)

This article reviews the performance measurements from chapter 10 of my book Optimized C++, plus a new type of container, the circular buffer, that is a candidate for inclusion in C++20.

Longer Works

Technical Report on C++ Performance ISO/IEC TR 18015:2004 (PDF)

This is a definitive 202 page report on performance-related issues in C++03 by the ISO C++ Committee, fully supported with extensive timing evidence. At the time this TR was written, C++ was under an unjustified attack for being inefficient compared to C. TR 18015 is amazing, offering a mature look at C++ which includes much of the optimization advice in Optimized C++, albeit in a compressed form. I was aware of TR 18015 before I began writing my book,  then forgot about it until the book was in production.It's a gem.

TR 18015 covers all of the following:

• Types of programs (embedded programs on small devices, servers running flat out) with performance issues
• Costs of various C++ features: namespaces, type conversion and casting, classes and member function calls, exception handling, and templates, all supported with run time measurements.
• Two methods of exception handling (code, table). This discussion also contains a list of alternative error handling mechanisms.
• Ways to make the standard I/O functions faster by fiddling with locales.
• A hardware interface using atomic and volatile. 

TR 18015 spawned some articles in the embedded systems press. Some of these articles offer a much less mature “C++ is less efficient than C” viewpoint.

Efficient C/C++ Coding Techniques, Moss, 2000

"Explains" that C++ is less efficient than C if you use virtual multiple inheritance when simple containment will do. Some comparisons of control flow choices also motivated by poor understanding about what’s good about C++. Great diagrams on virtual functions and virtual inheritance, though.

Using C++ Efficiently In Embedded Applications, Quiroz, 1998

"Explains" that C++ exception handling is expensive because you don’t really need to destroy objects as you exit each stack frame, which is fine if resource leaks are OK with you. Paper talks about some C++ features and categorizes them into cheap, maybe-cheap, and expensive categories.

Reducing Run-Time Overhead in C++ Programs, Saks, 2002

A paper on optimizing C++. It has great 90/10 law quotes, and covers implicit copying and custom new and delete. It cites Bulka and Mayhew in its bibliography.

Representing and Manipulating Hardware in Standard C and C++, Saks

A paper on how hardware registers may be represented in C++.

Optimizing Software in C++: Agner Fog (PDF)

This is a long (150 page) treatment of statement and expression-level optimization in C++. The material is informed by deep knowledge of the x86 architecture, and makes reference to architectural details. It is only partially translatable to other platforms. Agner Fog actively maintains this document, so its content changes from time to time. If you find it useful, it may be worth caching a copy for your own use, lest the material change out from under you.

Optimizing C++: A book about improving C++ performance (Wikipedia | PDF)

There is some useful information in this small collection of Wikipedia entries, though not enough to call it a book. Good things about this wikibook include:

• It evolved substantially during the time I was writing Optimized C++, growing to cover more material better.
• It has a bibliography, including free on-line references, and some books that were new to me.

There are several problems with it, though:

• No topic is covered exhaustively. There are, for instance, a few algorithmic techniques, but not a complete or even extensive catalog.
• Most optimization advice is supported only by hand-waving arguments about how a compiler might behave.

Books

Efficient C++, Dov Bulka and David Mayhew (Amazon link) 1999, Addison Wesley, ISBN: 0-201-37950-3

Meticulously researched, but now slightly dated. I have tremendous respect for this work.

The Software Optimization Cookbook, Gerber, Bik, Smith, Tian (PDF table of contents)

I have not read this book, but it appears to be a useful source of information on processor-dependent optimization for the IA-32 (Intel 32-bit) architecture.

Optimizing C++, Steve Heller (HTML), 1999, Prentice-Hall, ISBN: 0-13-977430-0

Warning: not about performance tuning in C++ at all, but rather about optimizing an ISAM database. You can view it for free on the web.

Optimized C++, Kurt Guntheroth (Amazon link), 2016, O'Reilly Media, ISBN: 978-1491922064

My book, the best of everything, of course.

Other Notes

Microsoft Visual C++ Tips and Tricks has all sorts of debugger tricks. Not about optimization, but interesting.

Mastery Takes Time

I have talked about achieving mastery previously. Here's another, perhaps better researched article on how long it takes to achieve mastery of a discipline. For those too indolent to click a link, what it says is that it takes about 10,000 hours, or about five full-time years, to obtain a reasonable degree of mastery, and it doesn't matter what it is; playing the violin, swinging a baseball bat, or coding in C++. A study of violinists indicated that "natural" talent was not a factor. The people who were the best were the ones who practiced the most and the longest.

This is something recent college grads all seem not to know about the careers on which they are embarking. They think (like I once thought) that their natural intelligence and their good education was what made the difference. As an Old Hand, I can tell you that practice has a lot to do with it, and good as you are, you will get a lot better when you've been doing the job for 20 years.

Don't Teach Yourself to Code by Writing a Library

It is a terrible idea to teach yourself how to use some unfamiliar feature of C++ by designing a library that heavily uses this feature.

I cannot think of an easily explained example, because libraries that come into existence in this way are very rarely easily explained. Even experienced C++ folks look at the thing and say, "Huh...WHAT?! Can you DO THAT?" Readers with more than a couple years of development experience are grimacing right now, because chances are good that the code they are paid to work in every day relies on a library that some genius thought would be cooler if it all used some tricks they'd been reading about.

It may be cool to the original author, but to anyone who hasn't read the same blogs as the author, it's just mysterious. Especially mysterious when the author has trouble doing what they originally imagined was possible, and ends up covering this lack of experience with a hack, that then lingers long after the author stops working on the library.

Writing libraries is a great time to slow down; to be conservative; to solicit the widest possible review; to ask your peers, "Is this good style?"

The Mysterious Hidden Languages of C++

C++ is old enough to have teenaged children (like Java, conceived when C++ was itself a wild teenager). Some lucky developers have spent virtually their whole career writing C++. But while C++ grew and solidified, the world of programming languages bubbled and foamed around it. Languages and language ideas came and went. Lexical styles changed and changed again. In these modern days of Hascell, python and Javascript, it can be hard to remember that the folks who originally designed C and C++ learned to program in BCPL.
 

The Secret Expression Language Inside C++

An expression language is a language where every executable statement produces a value. The vestigal expression language embedded in C++ includes individual expressions, sequences, blocks, if/else, and function call, but not iteration, alas.

x = 1, y = 2, z = x + y is a sequence. It sets the values of variables x, y, and z, and produces the value 3.

(rand(), 0) is a block. It calls the function rand(), and produces the value 0.

x == 1 ? y : (z, 0) is an if/else statement. It evaluates the boolean condition x == 1. If x and y have the values set in the previous expression, it produces the value 2. If x is not equal to 1, it produces zero.
 

C++ Has a Shiny New Functional Language Too

Here's the old syntax for a function.

int foo(int x)
{
    return x*x-1;
}

Here's the same function as a lambda. Return value in a different place. Syntax for associating a name with the function different. New rules to remember about the difference between lambdas and function pointers.

auto foo = [](int x)
{
    return x*x-1;
};

Lambdas have a whole new scheme (ouch, pun) of type deduction rules. Someday they are going to be special. But today they are mostly syntactic sugar, for people who grew up with Javascript or some exotic functional language so they can pretend that C++ is trendy too.

Aphorisms for Work

Here are some aphorisms that I think upon, usually during meetings with grumpy managers.

It takes the same time to do a project right as to do it wrong

You just spend the tine in different places. You can rush through coding without any planning, but you will spend a lot of time in front of the debugger, and maybe have to throw your code away and start over when you understand the problem better. Or you can do thoughtful design, code at a deliberate pace, run your unit tests, and be done. It comes down to a lifestyle choice.

The most dangerous thing in the world is a half-understood Harvard Business Review article 

There's an HBR article that says, essentially, "All other things being equal, getting software done sooner results in greater revenue." This one paper is the cause of so much misery for development teams. Managers love to use this article as an excuse to pry the software out of development and jam it into production before it's finished. They never read the "all other things being equal" part of the article. 

There is an HBR article that says, "Pay is not what motivates employees the most." There are managers who read this article and think they can overwork and underpay their employees. They don't read the part about what does motivate employees, like empowerment, recognition, and work/life balance.

The most pain is inflicted when the manager hears a summary second-hand, and doesn't even know he has only half the advice. He thinks, "There's an HRB article that says you should ship the project as early as possible", or "HBR says employees don't care that much about pay."

When everything is important, nothing is

Ever have your manager tell you on Monday to work on this, because it is the most important thing, and then tell you on Tuesday to work on that, because it is the most important thing. And then you ask him to choose which is most important, and he says they both are? That is what a weak manager does when he doesn't want to make a decision. There must be a most important thing, the thing to finish first. 

If everything is important, if everything is an emergency, then nothing is really important, because your manager will blame you for what you didn't choose, no matter what you did choose. If your manager behaves in this way, it's a good time to get your resume updated and start looking for work.

The 90/90 Rule: The first 90% of development takes 90% of the time, and the last 10% takes the other 90% of the time

Oh, how many times has a developer said he was 90% done in one month, only to spend a second month on that last pesky ten percent? You're 90 percent done with the part of the task you know about. Generally that's about half the task. It's a very mature organization indeed that can correctly estimate completion percentages. Your best chance is to estimate things using a consistent process, and then consistently record how your estimate matched reality. Then you can de-rate future estimates until your estimation process matures.

Software Quality: A Race to the Bottom?

Another Old Hand, whose opinions are based on very relevant experience, says that the software industry is trapped in a race to the bottom in software quality.

The trap is that investors' needs are divorced from software companies' mechanisms for producing value. Investors want the stock price or corporate valuation to improve in every reporting period; every year, every quarter, every month, every day. Investors are entirely uninterested in what a company did last month. They invest today to reap profit in the future. Companies, on the other hand, can only demonstrate their ability to produce value by completing projects; shipping disks, posting content, publishing interfaces. But it is hard to encode much value into a software product in a day or a month.

To impress investors, a company must ship, soon. Since investors aren't interested in what the software does, there is a strong incentive to ship early, but less incentive to ship complete or useful code. For a startup which has never shipped, it's all about who has the value story with the most zeros on the end of it and the shortest time to delivery.

To impress partners and customers, a company must ship. Customers are largely uninterested in what a company plans to do. They want products now. If a potential customer wants functionality, and there are competitors, the one who ships first wins the customer. Having won, it's a relatively smaller problem to keep the customer. Customers are used to defective software.

Thus, according to my friend, what wins is crazy, caffeine-fueled speed. Quality doesn't matter. By the time the customer finds out the product is imperfect, the sale is made, the money paid.

This reasoning depresses me. I hate what it means for my profession. Yet I can find no flaws in the argument, only small exceptions; established quality brands; software products (databases, kernels) that can never work in the marketplace if they aren't reliable; special customers who know what they need and write quality into their purchase agrements.

Reading Your Own Press Releases: A Management Antipattern

To customers, a new company is an unknown. A new company has no reputation. A new company must focus on providing innovative or cost-effective solutions to customer problems, because it's the only story they have to tell. Nothing wrong with that.

Well-established companies have more to say. In addition to presenting their latest product's features, they can tell customers about their previous products, their many years in business, their reputation for quality. Nothing wrong with that either. It's easy for a customer to choose an established company versus an upstart competitor, even without doing a bunch of homework, because the established company is a known quantity.

But sometimes, an established company loses its way. They begin to tell themselves about their reputation for quality, their years in business, and their leadership position. Their executives tell their product teams that they can charge a premium price because they are the "Cadillac brand", regardless of the strength of competitive products. If they say this often enough, it becomes received wisdom within the company. I call this behavior reading your own press releases.

A company that is reading its own press releases is already past its prime. When a company starts reading its own press releases, it stops worrying about solving customer problems, or being better, faster, and stronger than the competition. They may even get away with this for a few years if they really were market leaders. But eventually, they find themselves sidelined by their hungrier competitors, and their position of advantage evaporates.

I've worked for two companies who had this problem. One is now a division of a conglomerate, who fixed the problem by replacing complacent managers. The other annoyed its partners so much with slow product delivery that the partners made the company irrelevant. Last I checked, this second company was still in business, but they shed three quarters of their staff in a paroxysm of downsizing, and never recovered their former greatness.

I bet you can see some obvious examples in the news of large technology companies who have been reading their own press releases. Where are their new products? Missing in action or underwhelming critics and customers. When your division manager starts talking this way, tell them to wake up and smell the innovation.

Smartest Guy in the Room: An Organizational Anti-Pattern

Every software developer thinks they're smart. Heck, I think I'm very smart. But organizations seem to go badly wrong any time they anoint one person (or a few people) as the guy(s) to make all the smart decisions, and demote everyone else to code monkey.

I worked in a big organization where several senior developers convinced management to take them out of their business units, and put them in a special R&D cost center. They were going to do user interface research that would inform the next generation of products. Now, if this had been Bell Labs, and these guys had all been PhDs, they might have discovered the Cosmic Microwave Background or something. But these were just guys. When the team I was on asked these special devs to do some user interface research for our next product, they said, "Oh no, we can't be tied to a particular product." As far as I know this group never produced so much as a PowerPoint presentation. They were disbanded after a couple of years.

A company where I worked purchased a startup that found itself out of money and on the market for cheap. Their fast-talking CTO was a former employee of my company. He talked my company into putting him in charge of all development in our division, singing the following siren song. "You've been spending way too much time and money on carefully designing and architecting your products. At my company, what we did was just built 'things' and put them in a pile. Once a year or so, I would reach into the pile and pull out enough 'things' to make a shippable product." This strategy had worked so successfully for his last two companies that one had been folded into the other, and the survivor sold to us. It worked so well for us that he went through two years and six million dollars without producing a single product. Our division was shut down and everyone laid off, including the fast-talking smart guy, but not the two managers responsible for hiring him.

At another company, a smart team was created to write a big Windows editor to solve our very complicated configuration problem for a variety of products. They spent 18 months working (not too successfully) on a universal data input editor. But they never considered the complex problem of compiling the input data for each of several rather different products into usable files. They had in fact no idea how to solve this problem. So they never showed a single working version. They built a rather elaborate castle in the air, starting from the fancy gabled roof downward, hoping they would have an inspiration for how to build the foundation when they got to that point.

Moral: The smartest guy in the room needs to be managed just as much as the most ordinary. Otherwise they are liable to go off course and waste a bunch of time and talent, just like anyone else.

NASA's "Decline" is America's Triumph

On the day I was born, the only thing orbiting the Earth was the moon. By the time I was 5, however, the two most advanced nation-states had gotten men into orbit. This culminated, when I was twelve, in a manned moon landing. I recall very clearly how the entire world seemed to stop and take a collective breath as terrible, grainy TV pictures and prosaic radio traffic chronicled mankind's greatest technical achievement.

Fifty years on, only one more nation has launched men into orbit, and a coalition of european nations can send hardware into deep space. But the United States no longer launches men and women into orbit. We retired the Shuttle fleet after demonstrating that this task, while risky and expensive, could be performed routinely.

NASA has lost its way, some people say. America has lost its will, some people say. Another symbol of decline, say some.

I don't think so.

An American company called SpaceX, with only 1,100 employees, has learned to reliably launch payloads to orbit. Manned launches are close, held back mostly by concerns that their safety be more thoroughly demonstrated. A couple of other companies are close. Their cost per kilo to orbit is a fraction of that charged by the defense giants who assemble rockets for governments.

This result is astonishing. It means, here in the U.S. at least, we don't need the government to fund commercial activities in space. They fund themselves. It is a triumph of capitalism, almost unheralded. What it means is, the United States wasn't sitting on its hands for the last fifty years. We were investing in materials science, electronics, software, and all the things that actually make up a space mission. We were making launch capability reliable enough that companies would risk nine-digit bankrolls on activities in space. In the language of business, we built a reliable supply chain to space. This was something NASA knew we needed to do way back in 1980.

And where is NASA? Right where they should be, reaching out for another planet that we cannot yet grasp. Probably all other planets by the time we put boots on Mars. This year (2014), there isn't one part of a Mars mission that we really know how to do; keeping human beings healthy and strong on a long voyage; soft-landing a big payload on Mars; building a rocket on-site for the return to Earth. Some of these problems can be solved by brute force and expenditure of treasure. That's what NASA is for. And right behind are the entrepeneurs and their engineering teams, looking to finesse these problems to make it economical.

To those who say we should forget about space and spend NASA's budget to end poverty and hunger, I have this to say. Poverty exists by definition. It can't be eradicated in a society that rewards achievement. And NASA's budget could not banish hunger from the world for as much as a single day, especially where and when evil men wield hunger as a political weapon. President Kennedy had it just right when he said we do things like exploring space, "not because they are easy, but because they are hard." Mankind needs something to strive against, in order to make progress. What Kennedy didn't mention was that the problem of going to Mars is actually more tractable than the problem of ending hunger.

And to those who say manned exploration is frivolously expensive when robots can do it better, I say this. Who cares if there is an ocean on Titan if we will never stand on its shore? Why return pictures of Mars' dusty, cobbled plains if we will never walk them? Expanding mankind's reach is the only reason to visit space. We should only use robots where we cannot yet venture.

Mankind's conquest of orbit is nearly complete. The geeky men and women who will design, build, and fly the first manned Mars mission are in high school right now. They grew up in a world where manned space launches were so routine they only made the news by being so dramatically photogenic.

I expect to watch the Mars landing in beautiful hi-def color before I die of old age. Maybe on my wall-sized flat screen, maybe on my wristwatch. The men who landed on the moon were my father's age. The men and women who land on Mars will be my childrens' age.

So stop wasting time reading this and get to work!

One-Man Code Review

How do you do a code review if you are working alone? You obviously have already read your own code and found it beautiful. Code review fits into the development process as one means of getting feedback on the quality of the code you are writing. Basically, you increase use of all other tools that give the same kind of feedback.

• Turn the warning level of your compiler all the way up. Do a release build if you're using Visual C++ because the release build reports more errors.
• Install and run a static checker. CppCheck is a free static analysis tool. Coverity is a paid tool. LLVM has a static checker, or so I hear. If you have an expensive version of Visual Studio, there are static checkers, but I've never gotten to use one.
• Memory checking tools that find frees of invalid or already-free storage are built into visual studio and available for free for linux, and there's always valgrind if you're patient. 
• Write and run a set of module tests.
• Profiling tools produce results that may cause you to rethink your interfaces. Writing module tests forces you to use the API's you create, which often causes you to change things.

We're Special -- a Conversational Anti-Pattern

I've worked within organizations that told themselves that some best practice widely promoted throughout the industry cannot possibly be applied at this organization, because this organization's work is "special". Our code isn't like other code. Our customers aren't like other customers. Our problems aren't like other companies problems, so this commonsense rule does not apply to us (and therefore we need not change our practice).

When I had the evangelical furvor of the rookie, this behavior made me wild. It was clear to me that there was absolutely nothing special about our code, and that industry best practices got to be best practices exactly because they were widely applicable. I came to understand that the purpose of the word "special" in this conversation was to defeat any kind of reasoned argument, any comparison or evidence, and defend the status quo. 

When I became an old hand (and by that I mean, as I was typing this blog entry after facing this anti-pattern yet again), I realized something else. I realized that "special" was the antidote to the phrase "best practice", which itself is meant to defeat argument without presenting evidence.

When you find yourself in such a conversation, it means that both sides have staked out a position that they are too lazy to defend with any actual evidence. To make any progress you must change the conversation. Perhaps, "Best practice X will improve our results in thus-and-such a way.", or, "Practice P is being used successfully by our own competitors." This raises the level of the conversation so that the other stakeholders must argue specifically against your evidence. When they say "special", you can then say, "Yes, but in what way does our special-ness invalidate the evidence?"

It is tedious to argue from evidence, and easy to resort to words like "special" and "best practice". However, the comversation generally goes better if you come prepared with some actual data.

How Code Review Fits Into the Development Process

Code review needs to be understood as a single tool in a toolbox that hopefully contains many tools that give feedback on the quality of the software development process. Which tool is most appropriate in a given situation depends on how much time and money are available for removing defects, and how much each remaining defect costs. Many tools point to defects in development

• code review
• linters, static checkers, and compiler error messages
• test suites
• style checkers

Code review is very good at finding weaknesses in the way individual developers understand their programming language tool. Developers who get busted even once or twice for (for instance) not destroying dynamic data or not initializing data structures quickly internalize these skills.

Human code review is good at finding coding errors that escaped the compiler, but static checkers also find some of these errors. Static checkers have little up-front cost and negligable cost to re-run, but their coverage is limited. By contrast, human code review is expensive each time you do it, but they have as much coverage as you can afford.

As a human process, code review is not well suited to reveal defects like performance issues, that require both a specification of acceptable performance and an accurate measurement of a running system. Human processes are also expensive to use for checking conformance with style guidelines, for which fully automated tools are generally available. This doesn't mean you can't use code review to find these defects. It just means there may be more suitable tools.

Human review of interfaces (for instance C++ header files) is good at finding squishy issues that automated tools aren't any good at finding. There isn't an automated tool yet that can judge the quality of a class or library API. A bad API isn't *wrong* per se. It's just more confused or more difficult to use than it needs to be. Class level APIs aren't captured as user requirements, but evolve out of design and use. Test cases developed early in the class design are also good at detecting API issues. The test writer goes to test a feature, and finds they can't get at the feature in a useful way. The advantage of test cases is that they are inexpensive to re-run if the API implementation changes, while code review is expensive every time you use it.

Pair programming is the ultimate in code review. While one person types code, another reads the code and makes suggestions. It's also the most expensive form of review, costing a man hour of review for each man hour spent coding. I think code review meetings have almost the same coverage, but at a far lower cost.

There are automated tools that seek to make code review efficient. You can get your source control system to hold off checkins until people have signaled approval of the code. I find these tools are less effective than a code review that involves an actual face-to-face meeting. When each developer reviews in isolation, it's too easy to get lazy and perform a prefunctory read of the changed lines before signing off on the review. I used a code review process that required face-to-face meetings among three or more participants. I felt the cost was justified by the quality of the review.

How I Do a Code Review

I led a team to do a design review of every base class in a 40-man-year project; something like 200 C++ class definition .h files. If there was code in the corresponding .cpp file, we sometimes reviewed that too, but we performed this review early in the design phase, so there was often only a header file of 100-200 lines.

We agreed as a team to follow good-meeting rules. Every design review meeting required a minimum of 24 hours notice, and the review materials also had to be complete and distributed 24 hours in advance. If the materials were not ready in time, the meeting was rescheduled. Each participant was required to have reviewed the materials in advance, and made notes. At the beginning of the meeting, the facilitator asked if everybody had done their homework. If there were not at least three people who had come prepared, the meeting was terminated and rescheduled. Meetings were held face-to-face in a conference room, were scheduled for 60 minutes, started on time, and were stopped and rescheduled if they ran over. In this way, we ensured that the client (the person whose class was being reviewed) had prepared, and that a high quality discussion would justify the cost of pulling three or more people away from work for an hour.

A facilitator other than the client was selected for each meeting. The facilitator ensured the meeting stayed on topic, took notes on issues discussed, and circulated the notes to the participants after the meeting. After the meeting, the client edited the notes to indicate what action they'd taken on each issue. The notes were archived with the other project files.

The materials prepared for each meeting were the class's .h file listing, printed with line numbers for easy referencing, the class's CRC card (a concise and structured form of design document), and any diagrams or other materials the client considered relevant.

Since I knew that having your code reviewed could be ego-busting, and because I was a relatively experienced C++ developer, I became the client of the first two or three review meetings. One point of these first meetings was to demonstrate that even an experienced developer could benefit from a design review, and the team did indeed find issues in my lovely code that needed addressing.

When the inevitable snarky comments started coming in that first meeting, I reminded the team, "Today it's my turn. Next week, it'll be you. Lets keep the tone professional, because you want it to be professional when it's your turn in the spotlight." This set the tone for all the meetings to follow. I don't believe we had a single incident of hurt feelings over the whole project.

After the first couple of meetings, it became clear that there were standard questions we would ask about every class. I compiled a list of these questions and archived it as a project document. (They read like Cliff Notes for Meyers' Effective C++). Within a couple of weeks, these issues disappeared from classes entering the review process.

The cost of preparation, and the keenness of team questions pushed every developer to produce high quality artifacts for review. This had a corresponding positive impact on the quality of subsequent code. Questions about destroying allocated objects led the whole team to take up the use of smart pointers. The RAII idiom was demonstrated and became commonplace in the code. In this way, the best developers taught good practice to the rest of the team, who were learning C++ as they wrote the code.

Later on, as the project approached code complete, we began to review the code, using the same meeting process. These code reviews captured many defects that had escaped testing. Our intention was to review only the most critical code, nominated by individual developers. By the end of the project, we had covered about 60% of the code with at least one review.

The software for this project achieved all its design goals and delivered high initial quality. During the development, our company decided to do ISO 9000. This project received the highest marks from the auditors over a period of a couple of years. Subjectively, we all agreed that we delivered significantly higher quality work than previous projects.

One important aspect of the review process was the team-oriented, consensual model we adopted. Developers were not required to remedy every issue raised at reviews. There was no sign-off or other coercive process in place. The team came to respect the review process precicely because it removed defects from their code. It also respected their individual judgement, both as coders and as reviewers.

The cost of the reviews was significant. My own personal experience was that I spent several days after I was "finished" with a class, getting it ready for review, getting the documentation complete, printing listings, etc. On balance, I thought the cost was justified. I would use the same process again.

Things Only Taught by Time

What do you learn over a career of coding, besides 123 different APIs? What's all this value in being an Old Hand? Well...

• When I was interviewing for work, right out of college, there was a job I didn't take because all the developers had foot-thick listings on top of their filing cabinets. I knew my modest brain could never comprehend such a massive amount of code. Ten years later, I delayed learning Windows programming because applications appeared indecipherably complex, with dlls and a configuration database instead of a single executable.
  
  I don't print paper listings any longer, but if I did, mine would be ten feet tall. I've built many Windows apps. It sucks to manage dlls, but not because they're too complex to comprehend.
  
  LESSON: Anything that other programmers are comfortable with is not too complex for you.
  

• I spent 10 years becoming a deep domain expert at my first company out of college. I figured I would become a "lifer", safe from the travails of the job market because of my valuable knowledge. Then during an economic downturn, the company changed strategic direction. They shuttered my whole division, laying me off along with 19 of 21 engineers, 12 of 12 marketing folks, and about 50 factory workers.
  
  It was hard finding a new job because (1) it was the bottom of a recession, (2) my deep domain knowledge was not applicable at any other employer in the city, and (3) I had neglected to learn the latest programming skills because I didn't believe, as a lifer, that I would need them to be up-to-date.
  
  LESSON: Skills and experience are only valuable if they help you find work.
  
  LESSON: Skills and experience that help you find work are valuable.
  

• The two engineers that my one-time lifetime employer retained? One was a very lucky new college hire. The company valued its reputation for firm job offers. The other survivor was a nice lady; very quiet, someone who never asked questions or made requests. She wasn't the smartest engineer. She wasn't the most innovative. But she turned out an utterly reliable so-many-lines of code each month without variation. She became a lifer at that company.
  
  LESSON: what managers value in a software developer is not intelligence, or innovation, or great code. They value reliably low maintenance workers.
  

• My very favorite sister-in-law died of cancer at 39 years of age. When her cancer was diagnosed, she didn't quit her job right away, but she did change her behavior. All of a sudden there were some meetings and some tasks that seemed so unnecessary that they offended her sense of limited remaining time. She told me that a week before her diagnosis, she had wasted a bunch of time in these meetings, and all of a sudden she really wanted that time back. She began focusing exclusively on the parts of her job that added value and that gave her satisfaction. She put off the dumb stuff as long as she could. She discovered that lots of dumb stuff eventually just went away if she put it off, because it was dumb stuff. She didn't lose touch by skipping boring meetings. They were boring because nothing happened in them. In this way she became recognized by her managers as the most productive worker in her office.
  
  I was going to boring meetings where nothing happened too. I vowed to behave as though I had six months to live. I became more productive, because I only did activities that added real value, and that I enjoyed. I came to feel as though I had become bulletproof. I reasoned that if I was ever let go for only doing productive work, the company would be doing me a tremendous favor. Such a company would not last long, and working there until the end would be awful.
  
  LESSON: Only do tasks that add value. You will be the most productive member of your team.

• I became unemployed in the dot.com crash, becaue my dot.com employer collapsed. It was hard to find work because it was the dot.com crash, and nobody was hiring. It wasn't that I had the wrong skills or wanted too much money. There just wasn't anyplace to send a resume. It turned out that I wasn't bulletproof after all. I went from being the most employable guy I knew to being chronically unemployed. I became unemployed again in the Great Recession. Two employers in a row suffered dramatic, thirty per-cent revenue declines, and laid off their whole software team.
  
  LESSON: Bulletproof is not the same as invulnerable. Pride goeth before a fall.
  
  LESSON: Most software development work is project-oriented. Your job is always vulnerable between projects.
  
  LESSON: It is always the bottom of a recession when you get laid off. Nobody will be starting new projects then. It is prudent to have money in the bank to last you a year or so.
  

Looking back on these lessons, they seem very obvious. But each one had to be learned. They were all novel thoughts until after they happened. I'm not sure it's even possible for a rookie to internalize these lessons because the "duh" only comes with experience.

Time to Change Jobs?

It sucks to be laid off. On top of the emotional response to being rejected and discarded, you have to begin a job hunt from a standing start, while simultaneously finding ways to keep food on the table. It's better for both your ego and your wallet if you pick the day of your departure and move directly to a new job.

When is it time to change jobs?

Change jobs ASAP if you think the management is leading the company in the wrong direction, if they are mistreating you or other employees, or if they want you to do something illegal or unethical. There is no future in such a job. Even if you stay, you'll either get fired (because managers can tell when you think they're leading the company to trouble), or the company will fold (because the managers led the company to trouble).

Change jobs if your employer is working you 60 hours every week and they aren't paying you a salary-and-a-half. There are lots of companies where employees work a reasonable number of hours. Most companies that make you work long hours don't pay any better than average. Why give them your life for nothing? Yeah, every job has crunch time when a deadline looms and you gotta put in the hours. I'm not talking about that. I'm talking about companies who think all developers should work 60 hours a week because the founder worked 60 hours a week at Microsoft back before the IPO and got rich on stock options. Stock options are not all created equal. Most expire worthless.

Change jobs if you aren't challenged by what you're doing, and you like to be challenged. HOWEVER, be mindful that you will continue to discover new and better ways to use whatever tools you're using for five years or more. Folks just out of school very often think they know all they will ever need to know by the time they are 25. This is so totally not true, but you won't realize it until you're 35 or 40. Find an old hand you respect and ask them. They will say the same thing. You will just have to trust them until you feel it for yourself.

Don't change jobs if you are really happy where you are. Happy matters a lot more than rich. That's my advice. HOWEVER, don't confuse mere comfort with happiness. It's easy to get comfortable, thinking you know everything in the world about some mature product written in some old language, and that you can set the cruise control and be employed for life. But your company may suddenly drop that mature product, and then where will you be with your out-of-date skills and no relevant industry experience?

Change jobs if what you're doing at this job won't help you get another job. I interviewed a software engineer once who had never written any software. He administered contracts of the third-party companies who wrote tests for the software written by other third-party companies. He'd been in this job long enough to forget his CS classes, and had gained no relevant experience. Needless to say, he didn't get the job. I sat near a software engineer who wrote code for a ridiculously small microprocessor for embedded devices. It was so small, its program counter was a feedback shift register, not a counter (fewer gates). Branches and jumps could only target every 16th storage location. There was no macro assember that could emit code for this processor, so you coded it in by hand in hexadecimal bytes. He worked at this job for five years before getting disgruntled, and found he had zero relevant experience anywhere else on the planet. I am personally one of the planet's leading experts on functional RAM testing. Guess how many jobs there are in that field.

Change jobs if you are being paid less than friends make for the same work. HOWEVER, don't change jobs because you heard that's the way to pump up your salary. If you are underpaid and have hot skills and are just a couple of years out of school, job-hopping might work for you once or twice. But in the long run, you can't pump your salary infinitely high. HR directors meet in groups or exchange emails or subscribe to reports that tell them what software developers with so-many years experience are making. The only things that will improve your salary in the long run are visible accomplishments, years of experience, and having a variety of very up-to-date skills.

A little humility goes a long way when thinking about changing jobs. Don't change jobs because you heard a story on the internet that some guy one year out of school is pulling down $150k. It doesn't mean that you can get that much (even if you believe stories you read on the internet). Freakishly rich companies like Facebook and Google do pay better than regular companies. Facebook and Google are correspondingly more fussy about who they hire. You may not make the cut, even if you think you're a smart guy.

What I Hate About Agile

Blasphemy, I know. Lest the keyboard burn my fingers, I should say that I really like Agile a lot. What I don't like about Agile is that it means whatever you choose it to mean. Agile is a brand name, useful for selling books and expensive consulting. But it isn't a specific methodology. Consultants and Agile boosters use the Agile brand the way terrorists use the al Qaeda brand, to make whatever agenda they happen to be pushing sound more powerful.

If a dev organization uses any Agile practices at all, and delivers software at the end, its Agile boosters or Agile consultants proclaim, "See, Agile works." If the organization uses Agile practices and the software is late, broken or unusable, they say, "We just weren't Agile enough!"

Because Agile doesn't mean anything, it's easy to defend any practice, no matter how lame, by calling it Agile. Don't like being held to a schedule commitment? There are agilistas out there saying, "You can't estimate schedules anyway so Agile doesn't do schedules." Don't like specs and planning documents? There are Agile boosters saying, "The code is the documentation." Rather jump straight to coding instead of doing design? There are Agile-branded supporters for that too.

Particularly disingenuous are allusions by agilists of a certain stripe to the Standish Group's CHAOS reports and the "Software Crisis". Standish makes money by selling an expensive annual report which perports to show that most software projects are delivered late and over budget (which mean the same thing). This particular strain of agilists proposes to solve the crisis by refusing to schedule. That might cut it on the web, but anywhere hardware must be manufactured or media ordered to coincide with release, refusing to schedule just won't satisfy stakeholders. These same agilistas like to compare their brand of Agile against a straw man comprising the worst habits of traditional development. This doublespeak put me off taking Agile seriously for years.

I am also leery of developers who claim to be agilists, but whose agenda is to discard any rigor or discipline in an existing software process in favor of blasting off in a blaze of coding glory. There are Agile processes that don't do up-front design, and others that don't do documentation, scheduling, or whatever. These devs take this as an excuse to do away with all these things, with predictably appalling results.

I am also suspicious of one-size-fits-all prescriptive agile processes like XP. Pair programming (a required practice in XP) is helpful, but very expensive. A mature team can be successful with a less expensive feedback method; code review or maybe static analysis. XP may be the perfect process for a particular team in a particular company. It's just not perfect for every team at every company.

Most Agile shops I've worked in had decayed Agile processes, announced with great fanfare by Agile boosters some years before I arrived, and since collapsed to a few residual practices like feature request sticky-notes on the wall, or daily stand-up meetings. But the "sprints" were ten week marathons, and the code only shambled out of QA into customer hands a couple times a year. I'm sorry, but this isn't Agile. Not capital-A Agile, and not little-a agile either.

I think this situation is sad, because if every dev team combined rapid iteration with feedback from design review and test, plus the amount of rigor in other processes that was appropriate to the quality desired from the work products, we would live in a world of far more beautiful software than what we see today. I don't know what to call this methodology. Sure it's Agile, but it's Agile in a specific way.

What I Like About Agile

As an Old Hand, I've developed software under both Agile and traditional process methodologies. The traditional process enumerated requirements, architected the system, and produced the code as sequential phases. Each phase produced a document or artifact for formal review. Since the skills needed for these three phases are actually somewhat different, the various phases could even be done by specialist teams.

What I liked about the traditional process was that it was thoughtful and disciplined. Reviews at the end of each step ensured that all stakeholders agreed that the step was complete and the project was still relevant before the project was allowed to progress.

Several things were uncomfortable about the traditional process.

• The great bulk of the work took place during the coding phase. There was not an obvious place to review the relevance of the whole project during coding.

• Too often, the team only discovered it had missed a requirement or failed to do some critical bit of design deep into the coding phase. The end-of-phase reviews were good at checking for incorrect requirements or design issues, but too often failed to catch missing ones. By the time problems were revealed in coding, the project had spent hundreds of man-hours on design or coding that had then to be discarded and re-done.

• The team could not adapt its practice to improve the project because feedback only arrived at the end of each phase. End-of-phase reviews only helped improve the next project, and only if the team stayed together.

In my humble opinion, the Agile revolution brought just two new thoughts to software design.

• Performing many small iterations of the requirements-design-code cycle instead of one big one makes feedback available earlier in the project. This gives the team a chance to improve their process to reduce cost, decrease uncertainty, and improve quality.

• If partially completed software has any functionality, it can be released and start to earn value at once.

A host of specific practices like feature cards, test driven development, or pair programming have arisen, and been packaged into name-branded methodologies like Extreme Programming. But specific practices are not what makes a project Agile. Frequent iteration, frequent release, and frequent feedback make a project Agile, in my opinion.

Agile projects do the same three tasks (requirements, design, coding) that traditional projects do, notwithstanding certain agile extremists who say they don't. A sound Agile process adopts practices that ensure that all these activities provide early feedback, and that the team acts upon this feedback.

Agile Practices Reviewed

I was so horrified by the waste I perceived in prescriptive agile methods like XP that I came very late to the agile party. Here are some specific Agile practices about which I have thoughts.

• Pair Programming: Pair Programming provides great feedback to individual developers on code quality, but it is very expensive. It's good for teaching inexperienced devs to code, but not so helpful when your team is already competent. Code review provides much the same benefit at lower cost.

• Code Review: There are automated tools for code review that present the user a visual diff of the modified code, and allow commenting and approval. While these tools are great for reviewing point-changes during maintenance, they discourage thorough review of whole interfaces. It's too easy sitting at your desk alone, to take a perfunctory glance at the changed lines, and say, "Looks good. Ship it." Shops using automated review tools have to keep an eye open to be sure all reviewers are taking a decent amount of time to actually read the code. I have had good results from a heavy-weight design review involving actual meetings. The formality and required prep work for the meetings, and the expectation of face-to-face feedback from peers induced higher quality even before the review.
  
  Linters and static checkers are tools you can use for code review too, even if you're only reviewing your own code.

• Unit Test: I love unit test. A good set of unit tests remove much risk from changing software. Unit tests written alongside (not after) developing the code form a check on the consistency and completeness of newly designed APIs. The needs of unit testing focus attention on separation of concerns and isolation of dependencies, which both push up the quality of the resulting code. Unit tests are most effective when devs buy into the idea of testing as you go. If writing the unit tests is just an annoying check-box to a developer, they are unlikely to give it the attention required for a really good result.

• Code as the Only Deliverable: Some agilists suggest that, since the code is the only artifact shipped to customers, no other artifact has any value. Production of requirements lists, design documents, schedules, and other non-executable artifacts should thus be viewed as wasteful.
  
  This advice has merit from an aspirational standpoint, but has practical weaknesses. Code is too voluminous and precise a language for expressing requirements. Code is too low-level to express architectural decisions. Code cannot express schedules at all. Other kinds of documents may be necessary for internal communication and review, even if they aren't delivered to customers.

• Schedules: There is a persistent myth that Agile methods don't do scheduling. This is untrue on many levels.
  
  At the micro level, the effort for individual features must be estimated, and big features broken up into sprint-sized pieces (if you're doing sprints. But otherwise you're doing big-bang development. Hisssss).
  
  At the macro level, Agile projects must do scheduling when stakeholders require it. This happens any time they interact with other projects, when long lead time hardware and mechanical packaging must be designed and ordered, or when the dev organization must release by a calendar date or forego important sales (in time for Christmas, for instance). Only a small subset of Agile projects can safely ignore macro scheduling issues, or refuse to predict completion.

• Refactoring: The world of software development is of two minds on the subject of refactoring. One school believes that any change to released software must be minimal, lest the change introduce bugs. The other school says refactoring is OK if it adds value by making the software more maintainable or more flexible, or better supports new features.
  
  In my opinion, both thoughts are schooled by experience. If the initial design and coding were weak, and there are no unit tests, then any change is risky, so change must be minimized. If, on the other hand, the initial design was well motivated and good unit tests are available, refactoring just makes things better and better.