The Perfect Info-Tech Workplace

There should be a kitchen with a lot of food, hot and cold beverages, etc.

I don’t necessarily agree that your workers should have spacious offices with doors that close, but they should still be comfortable enough. At one of my workplaces, I constantly had to move to let other people out of their seats, and this was unbearable. So don’t do that.

In one job that I fondly remember, every employee had their own spacious cube, with a desk to put a computer and a place to put a few chairs. This was much more like it.

In a different job, most developers worked in a spacious studio room over looking Kikar Hamedina. It was also comfortable to work there.

This cannot be stressed enough. As Joel Spolsky notes (based on Steve McConnell) in item No. 9 of the Joel Test, you need to “use the best tools that money can buy”.

If you buy old, broken and/or barely functioning hardware, you’ll spend a lot of time debugging the problems there, which will waste a lot of precious time. And you may lose a lot of reputation and customers due to down-time. Relying on reliable, high-end hardware is a much better idea.

I’ve been to two workplaces that gave me an old computer with a 40 GB hard-disk. It wasn’t enough at all. At one place, we’ve reached the limit of this hard-disk due to several large source code checkouts, and as a result needed a bigger hard-disk. And the only hard-disks the lab had were 80 GB ones, which were bought because they were the cheapest (per unit, not per capacity). Please, buy large enough hard-disks.

At the same workplace, I was given a computer with a read-only CD-ROM drive. It was not even a DVD reader! I brought a DVD of audio files from home, and could not read it. In this day and age, read/write DVD drives are the standard, and are ultra-cheap.

Sometimes you’ll need several computers, or a decent virtual machine emulator, to run alternative operating systems on the same machine.

Make sure your workers have a high-quality screen. They need to look at it most of the day, and they want it to look nice. A decent 19״ LCD screen nowadays is very cheap, and well worth the added productivity.

You also need a state-of-the-art version control system , of which there are currently several high-quality open-source alternatives. Some very costly version control systems have a bad reputation for being extreme troublemakers, while the modern open-source alternatives “just work”.

If there’s a good commercial software application that your employees like to use and can recommend, don’t hesitate to buy it. You’ll also find O’Reilly Safari Licences to be a good idea so your employees can easily look up and read information in many books on-line.

Give your employees sane working hours as Evan Robinson writes in “Why Crunch Mode Doesn’t Work: 6 Lessons”. I.e: 40 hours work-weeks or less. While sometimes asking them to stay late to meet a deadline is acceptable, the so-called “crunch mode” is under-effective and a recipe for disaster.

This feature was covered on Slashdot where there are many useful comments and later on a similar article (“Why We Have to Go Back to a 40-Hour Work Week to Keep Our Sanity”) was published in the site AlterNet, and also featured on Slashdot (with many comments).

This cannot be stressed enough. The more people work, the less their hourly throughput is going to be, until they end up at negative throughput. As a result, overworking workers is dangerous because you're risking introducing bugs and destroying existing work. Furthermore, your workers are going to be much less happy than if they worked less, which is also something you'd like to avoid.

Find a good place which is close enough for most people to get to. Make sure your programmers can commute to you easily. Even pay for taxi-cabs out of your own expense, or ask other employees to give them a ride, if there isn’t good public transportation. The money you spend on the cabs is very small compared to the one you would lose by lost productivity, and by frustrations of travelling.

Give your workers a lot of paid vacation. Humans are not machines - they need rest and relaxation. The 14 annual vacation days mandated by the Israeli government are a joke. You should give them much more than that. Joel Spolsky’s company - Fog Creek Software gives 6 weeks of paid vacation, which is much more reasonable.

Some organisations put their intranets behind firewalls that block access to almost all services. It is not uncommon to see only the HTTP and HTTPS ports open for free access.

However, there are many other Internet services that star programmers need in order to be productive. Among them are:

And naturally, malware can easily propagate and survive using HTTP and HTTPS alone, and there are ways that clueful workers can overcome such restrictions.

At one of my workplaces, I was able to chat on the IRC, connect using IM, ssh to everywhere I wanted, etc. without any restriction. It was a liberating feeling and I felt at home there. At a more recent one, I needed to connect to a remote host, and invoke port forwarding in order to talk on the IRC which was annoying and error prone.

So make sure your firewall is not over-zealous and does not prevent legitimate uses.

Competitive Salary? Yes, it’s important, but not absolutely everything. Great programmers often won’t work for you for free, albeit there are exceptions, like unpaid, volunteer, contributors to Red Hat’s Fedora project. However, they’ll find many other conditions much more tempting than an over-blown salary. See what Eric Raymond wrote about it in “Homesteading the Noosphere”, quoting many studies:

One final note: you might think to yourself: “how can I afford all that?”. Well, the answer is that the cost of all these advantages is very small in comparison to your general operation. And they will pay themselves much more as time goes by in the happiness, productivity and loyalty of your workers.

The last thing you want is to lose a competent developer. And all these recommendations will better make sure that he or she will stay with you.

The first advice I can give is to write functional specs. In the link, Joel Spolsky explains the motivation and method of writing functional spec. A functional spec describes how the end-user will use and interact with the program. It is not a technical spec which explains how the inner code will work.

From my experience, functional specs are fun to write, reveal many problems in the initial design, and make you think how the program will work on the inside.

Here are three examples for functional specs I wrote:

You should write automated tests such as unit tests, integration tests, system tests, that will run automatically on the code and yield an answer if all of them are successful, or if any of them fail.

There are many good practices for automated testing such as having daily builds or reaching a 100% test coverage. Here are some resources to get you started:

Why are automated tests important? For several important reasons:

In short, writing automated tests gives more confidence in the code and in changing it.

Note that having automated tests is not a substitute for having dedicated software testers (and vice versa). By all means, they are both necessary for any serious operation.

Most people agree that designing a program, planning it and thinking about it is a good idea. Extreme Programming suggests to have one design meeting every week, which sounds reasonable.

Joel Spolsky gives the motivation for designing software in one of his articles, which is a recommended read. He also voices an opinion about over-formal design (with lots of UML diagrams, etc.) that can actually be a disadvantage to a project’s success.

Refactoring is the process of improving the internal quality of the code (from “a big mess” to “squeaky-clean and modular code”), while not changing its external behaviour. This is done for mostly functional and bug-free code, but one which is not as elegant as it could be, so it can be better managed.

“Joel on Software” features two excellent articles about the motivation for refactoring: “Things You Should Never Do, Part I” (why rewriting functional code from scratch is a bad idea), and “Rub a dub dub” (how and why to do refactoring).

There are many other resources for that on-line, along with many refactoring patterns.

There are many types of refactoring: grand refactoring sessions (= what Joel describes), continuous refactoring (refactoring as you go), “just-in-time refactoring” (refactoring enough to achieve a certain task), etc.

But refactoring is important, makes development faster in the long run (and even in the short-run), and can prevent the code from deteriorating into an ugly, non-functional mess that would be hard to salvage.

You can also refer to my essay, and subsequent comments on it “Changing the Seldon Plan”:

There are a few software engineering methods that I find pointless, so I’d like to briefly point them out.

The first is the “Huge-design-up-front”, where an “architect” or even the entire team spend a very long time writing an extremely comprehensive and detailed technical document that specifies in detail how the software will work.

The problem with this approach is that it is a huge waste of time and also it is impossible to design a large project top-down like that. A better way is to involve the entire team in a good design session (a week long at most) while writing some functional specs, diagrams and some other documents, and then to write it incrementally.

A similar fallacy is the “Mountains of documentation fallacy” of having superfluous commenting, “Literate programming” (see the Wikipedia article), etc. The problem with this approach is that the extra documentation is often redundant if the code is well written and factored out ^{[ExtractMethod]}

Some documentation (especially API documentation such as Perl’s POD or Doxygen) is good, and you shouldn’t feel too guilty about writing a comment for an interesting trick, but too much documentation slows things down, and doesn’t help with the design process, and eventually may turn out to be misleading or harmful.

Something else I consider a bad idea is Overengineering or YAGNI (“You Ain’t Gonna Need It”). The basic idea is not to implement too many features at once, or over-complicate the design of the code, and instead focus on getting the necessary functionality working.

YAGNI, put aside, I still believe that some forward-planning is good.

function create_employee(params)
{
    let emp = Employee.new();

    emp.set_birth_year(params[year]);

    emp.set_experience_years(params[exp_amount]);

    emp.set_education(params[education]);

    ### Calculate the salary:
    emp.set_salary(emp.calc_education_factor(emp.get_education()) * emp.get_experience_years());

    ### More stuff snipped.

    return emp;
}

                        

The code-with-accompanied-documentation is somewhat a sub-case of a more general fallacy: the belief that a company can simultaneously maintain two different code-bases (say two code-bases implementing the same functionality in two different languages), while keeping them both synchronised.

The Extreme Programming experts have been warning against it, and it also makes sense that it is hard to do given that programmers, by nature, lack the appropriate discipline to keep maintaining two or more code-bases at once.

One possible solution to this problem has been illustrated by Fog Creek software. What they did was implement a compiler from a common language, to several target languages. Naturally, this compiler also needs to be maintained and extended, but it is less work and less error-prone than maintaining several different code-bases.

One of the anti-patterns I encountered in previous workplaces is that of the “Knowledge in the Ether”. Quoting from my original post:

There was also a comment there that said that if one encounters a repetitive task, they should write a script to do that. And it’s a good idea to keep good README files, comments, etc. and to use a standard building procedure, or prepare distribution-level packages for the program.

Nevertheless, a workplace should encourage its employees to note down every useful knowledge and procedure. My friend once told me that in his previous workplace they kept a knowledge-base as a group of Microsoft Word documents stored inside a Visual SourceSafe repository. As a result, people felt it was too complicated to update and maintain them. Eventually, he set up an instance of MediaWiki there, which proved to be much more convenient, accessible and quick.

Another essential program to use is a good web-based bug-tracker. Joel Spolsky has an article about the motivation and strategy for bug-tracking, which is well worth reading.

The reason you need it to be web-based is so that it would be accessible from all operating systems, so it can be exposed to the outside and so you wouldn’t need to install a client on all machines. Microsoft’s internal bug tracking system is not web-based (for legacy reasons) and they’ve been criticised for the fact that Internet Explorer, unlike the Mozilla-based browsers, does not have a publicly available bug-tracker.

Furthermore, in a previous workplace of mine, I was stationed on a Linux workstation (which was the deployment platform and what I like working on), and found that I couldn’t interact with our bug-tracker, which only had a windows client. Its more recent version already had a web-interface, but they wouldn’t upgrade for fear of breaking the other people’s workflow. This indicates another problem in not having a standards-compliant web-interface that can run on most browsers.

A good team should have a good version control server and clients available to every developer. I originally started the “Better SCM” site as a way to concentrate the material for the various alternative available, after I realised CVS has become inadequate for most needs.

My rule of the thumb for choosing a version control system is the following: “When in doubt - use Subversion”. See Ben Collins-Sussman’s article “Version Control and ‘the 80%’” for some of the motivation for this statement. Another one is that from my impression most of the new breed “Distributed Version Control Systems” tend to have catches for the unwary and other things that don’t exactly work as expected, and sometimes have poor portability to Microsoft Windows - a fact which is often a deal-breaker.

In this day and age, you’d probably expect the new system to be packaged for the target platforms (of which there can be many), to be capable of being configured according to the constraints of the platform and according to various limitations and to properly be able to be built incrementally and tested.

If you’re not convinced, refer to the Robert Jones’ OnLAMP.com article on “Planning for Disaster Recovery on LAMP Systems”, and What Joel Spolsky says in the Joel Test about “Making a build in one step”, and about having daily builds.

So you need a good building, configuration and packaging system. I covered such systems in an earlier article, and also maintain a list of links to some prominent ones and to other more comprehensive lists, so you have a high-quality selection.

But you need to use something, and preferably something common, standardised, well-tested and high-quality.

One huge advantage to a technology is that it will be open-source and available on several operating systems, including most UNIXes and Windows 32-bit, and on many CPU architectures. While, often you’ll encounter many platform-specific bugs or libraries or programs that work properly only on certain configurations (which is expected), you should still have the choice of being able to rely on the technology being present.

For example, Microsoft’s Visual Basic was a closed-source and Windows 32-bit and x86 specific programming language, which had proven to be very popular. However, Microsoft decided to discontinue it completely, and as a result, many of the applications developed for it can no longer be maintained into the future, and important bugs in Visual Basic will not get fixed.

Eric Raymond’s “Tale of J. Random Newbie” is illustrative of the many problems programmers encounter in highly-proprietary environments. However, even if you’re using proven, mature, well-documented and functional technology, you are likely to encounter bugs.

I like Perl a lot, and have recently found some bugs or even crashes in Perl code, which I’ve been trying to isolate and report. Perl is otherwise very good and reliable, but advanced users, or even intermediate ones are likely to discover many edge cases.

You should make sure you have the capability to isolate, report and possibly send a fix to such a technological problem, and get a prompt fix. This often depends on having good support from the technology vendor or developer.

Various technologies and development platforms, vary in their power and cost. As opposed to the possible misconception of “If you want something good, you’ll have to pay more.”, often cheaper, or even gratis (or open-source) solutions are more powerful (not to mention more reliable and faster) than their more costly counterparts.

Often, however, the really good solutions will be an overkill. Most experts I’ve talked with agreed that the Oracle Database is the most powerful SQL database on the market and “years ahead of anything else out there”. However, most run-of-the-mill web applications won’t make use of even a small percentage of its advanced features, enough to justify using it instead of, say, the open-source and gratis PostgreSQL.

For many problem domains, Oracle would make the most sense, but it’s still not very useful in the general case.

It makes sense to choose a technology which you know very well, so you can overcome problems more quickly. However, smart programmers can learn different technologies very quickly. As Spolsky notes:

If you expect that you will hire the best developers, as I pointed earlier, then they should probably be able to pick up something new fairly quickly. If, on the other hand, you’re going to hire bad programmers, then you’re likely to be screwed - even if they have extensive experience with their native technology.

So you should choose the best technology - not the one for which you expect to find the most “experts”.

Often a technology will get a bad reputation as being “quirky”, “hard-to-get-right”, “buggy”, “unreliable”, etc. This for example, has been the case against MySQL, against PHP, against Sendmail, and against other technologies. From my experience, such common criticisms often have a lot of substance, and should be evaluated, especially if there are equally affordable technologies, but such of a better reputation.

Maturity, Mind-share, and distance from the Mainstream are three factors that, like it or not, also affect the choice of language. Immature technologies tend to have many non-discoverable, or well-publicised quirks and bugs. Technologies without a good mind-share have fewer people to consult about problems one encounter. And the less mainstream a technology is, the less it tends to be polished or to have essential features, tools and APIs, that developers of more mainstream technologies take for granted.

As a result, such factors should be weighed in, when choosing a technology.

I once went to a job interview in downtown Tel-Aviv, which went very well. Then I received a phone call from them that they want me to come and install Fedora Linux on the computer. I did that, and next thing I knew, I was given other tasks.

I was instructed to write a mail-processing framework in PHP. Now since I know and love Perl, I know there are many fine modules for doing that on CPAN (= The Comprehensive Perl Archive Network), but there was little of substantial quality for PHP. At a certain time I needed to register at a site, to download the latest version of a PHP library, that was open-source software, because the download required authentication . There didn’t seem to be anything better.

I was told that they would prefer to write everything in PHP, because they expected to hire only PHP programmers, possibly without any Perl experience.

That wasn’t all. They also decided against using Postfix, which is a modern, high-performance and open-source SMTP server, and instead preferred the old Sendmail SMTP server which has a far worse reputation, and an arcane configuration system.

Furthermore, they also decided to use MySQL instead of PostgreSQL, and we ran into a PHP misbehaviour (which was fixable given a lot of PHP know-how) that made us have to restart the connection to the server after every request. Both MySQL and Sendmail were chosen due to political reasons.

With my PHP code barely working and prone to many errors, I decided to quit after about a month, out of being appalled by the bad code craftsmanship I had achieved there. Make sure you don’t repeat such a mistake.