Lessons From a JUG Talk With Eric ‘ESR’ Raymond

(full video of ESR’s presentation from YouTube below this post)

I have been President of the Philadelphia Area Java Users’ Group for 12 years, and as one might expect a typical meeting is geared around a presenter with a slide deck that gives a deep dive into some Java topic.  We could hear a case study, an explanation of a tool, tips for programming effectiveness, etc.   The JUG has followed this model since I founded the group in 2000, and we manage to get between 75-150 attendees at most meetings.

Last month we held a meeting that was entirely different.  I had reached out to Eric Raymond (aka ‘ESR’), who is best known as a leader in the open source software movement and author of The Cathedral and the Bazaar, about potentially speaking to the group.  Although ESR is not commonly associated with Java, I thought it would be an opportunity for the group to hear a well-known and respected engineer speak about a topic.  ESR suggested that he would do a free-form type presentation, without notes and slides, and simply take questions from the audience to use as improv material.  With about 150 engineers in attendance, ESR fielded a fairly wide array of questions on everything from functional languages to open source licensing to coding standards.

So why am I posting this article on JobTipsForGeeks?  As a recruiter I am often asked for career advice by my network of engineers, and my answers are always based much more on market trends (supply and demand) and ‘buzz’ around various technologies than on the viability of the technologies themselves.  For example, I am well qualified to discuss the adoption of specific languages by software companies in my region, but I am much less qualified to discuss the long-term viability of those adoption choices from a technical viewpoint.

As ESR was asked some of the same questions I am often asked, I thought his answers and insights were an opportunity for the audience to hear a ‘technically-grounded’ counterpoint (or supporting evidence) to the market-based advice I provide to individuals.  His specific commentary on functional programming languages and the value of learning them regardless of adoption rate was something I thought was insightful advice for engineers of all levels, and his musings on coding standards and how to obtain management ‘buy-in’ on open source are great tips for navigating at companies that may not be as tech-friendly as others.  These are not all specifically ‘job tips’, but I believe there is certainly some value in reading these opinions as you decide on which directions you may take your software career.

The quotes below are ESR’s responses to questions, organized by topic.

“Yes, I really like Python.  I like it for a very specific reason.  I like Python because of all the languages I have ever used, it is the one that maximizes ease of long term maintainability.  That is, the ease with which you can read your code six months later.  The longer I program, the more convinced I am that that is THE most important metric of a language, bar none… Most of the programming I do these days is in either Python or C.”

“I still haven’t actually learned Java well enough to do more than a couple hundred lines of programming in it.  I don’t dislike Java but I think it is a little over-verbose, it’s become kind of top heavy.  So it’s not my first choice, but if I had to write something in Java I wouldn’t go ‘ICK’.

ON C++
“C++ has the exact opposite problem to the virtue I had called out in Python.  Long-term maintainability of C++ code, TERRIBLE.”

ON PERL (answer continues from his answer on C++ above)
“OK, the best thing I can say about that is, it’s not as bad as Perl, but I’m afraid that constitutes damning with faint praise.  I still like Perl fine and occasionally use it, as long as the program isn’t more than 25 lines long.”

“I am sort of gingerly dipping my toes into the waters of Go, Google’s new language… I’ll tell you one concurrency thing I am really pleased by.  I have been wondering since about 1971 why nobody took the ball and ran with Hoare’s communicating sequential processes model.  So elegant, so pretty, so nice to reason about and 40 years later the Go people picked it up and ran with it.  That’s one reason I’m looking at Go.  CSP is the basis of their concurrency model in that language which is enough to motivate me to want to look at it some more.”

“I’m mostly ok with that.  I think the JVM has some deficiencies near word length and there are some serious problems with the numerical tower…It needs some work to be a really robust platform, it’s good but not as good as it needs to be.”

“I’ve dipped into Ruby a little bit, there was a point where I had to modify some Ruby code for a project I was working on, so I think I understand the language a little, maybe not master level.  My impression of Ruby is that it has pretty much the same virtues and the same problems as Python, and I might be tempted to switch except that it’s not different enough.  Functionally speaking of course, I mean aesthetically there is all kinds of odd little differences.  But it’s not different enough from Python to make me move, that’s my impression.”

“It’s on my list of languages to learn.  I have a friend whose judgment I trust who says it is a very good design, and that’s enough reason for me to go look at it.”

“…this is one of my more incendiary opinions, I don’t think we need the GPL anymore…My attitude in general is just use permissive licenses, stop with the viral stuff…”

“If you’re thinking in terms of bringing open source inside the corporation, you’ve already failed.  That is already thinking in the wrong direction because you’re trying to figure out how to control and make safe a process that thrives on lack of control and no safety, other than good code review.  Instead of thinking about how to bring open source inside the corporation, the right kind of question to ask is ‘how do I start up an open source project that benefits my corporation, build a community, and THEN sell it to my bosses?’.  Because one of the iron laws of dealing with bureaucracy is ‘it is easier to get forgiveness than permission’…How do I start an open source project that will interest my bosses, get it viable, and then sell it to them once they have a benefit that they can see?…Show them a benefit.  Don’t say ‘we can do this wonderful thing if you authorize me to spend n hours with no obvious physical return’.  That is something a manager is always going to say ‘no’ to.  What you have to do is show them a benefit.  It works much better, for example, to find an existing piece of open source software that is fairly close to solving the business problem and going to your boss and saying ‘you know, this thing already works and already has a user community and I can show you where it’s deployed, give me 50 hours and I can turn it into something that will solve our problem too’.  At that point you have a much better pitch, because that’s the kind of trade-off your manager is used to thinking about.”

“The universal argument that works for that is to say ‘hey boss, how would you like to reduce your maintenance costs?’.  Lay off the work on other people so it’s not coming out of your budget.  That’s the reason for open sourcing stuff that was developed in-house.  The argument for that is cost-spreading and risk-spreading.  And you want to put it exactly that way, ‘hey look, I’m going to reduce your bottom-line expenses’.”

“My first overarching observation is that the smartest thing you can do is choose a language in which coding standards are not necessary because there’s a uniform style that the language semi-enforces.  Yes I have Python in mind.  Go also has this property, it’s very difficult to have indent style variations in Go…The smartest thing you can do if all other things are equal is pick a language where you will never have coding standards wars.  The reason that that is a high-utility thing to do is, of course, the purpose of coding standards is to maximize readability of the code and long-term maintainability.  And yes, I do think that is extremely important… in particular if you are writing in C, the thing to do is pick a coding standard but don’t try and be too anal about it.  There comes a point at which the effort to enforce every conformance to every minutiae of a coding standard is causing you more pain and overhead than it is actually worth in reduced maintenance costs.”

“There’s a can of worms.  The first thing you need to know about me in this connection is I’m an old Lisphead…I actually cut my teeth on APL…The result of the first two languages that I learned is I have this mental measure of a programming language’s adhesiveness.  A programming language is adhesive to the degree that it sticks to your brain and cannot be displaced from your brain except by a language that is more adhesive.  I learned APL first, and then I got exposed to Lisp…My first question was ‘which one is more powerful in a practical sense?’, and yes, yes I know they are all Turing equivalent…So I decided to test the question by writing two toy implementations.  One of an APL interpreter in Lisp, and one of a Lisp interpreter in APL…That is why I switched to Lisp, and discovered that Lisp is more adhesive than APL, and it displaced APL from my brain.  Nothing has displaced Lisp from my brain.  I have not encountered any language that is more adhesive than Lisp.  Which is not to say I use Lisp a whole lot, but that it still dominates the way I think about programming.”

“I dipped my toes into Haskell, and someday I’ll have to do a significant project in Haskell.  But that time is not yet though.  I haven’t found anything for which it is better than the tools I’m already using.  The problem with languages like Haskell…and I have digested dozens and dozens of computer languages in my time and I used to be a mathematician with a specialty in formal and foundational logic.  With all these credentials, Haskell makes my brain hurt.  So if Haskell and other pure functional programming languages make your brain hurt, that’s ok.  And the thing that worries me about them, is that if they make MY brain hurt, how the hell are they going to make any traction with people who didn’t used to be foundational logicians with a specialization in logic.”

“My worry is that these are beautiful tools that will never actually achieve mass acceptance because they are simply too hard for most programmers to use.  The question is, what is the attraction then?  For the class of problems that is easily addressed using a functional language, using functional languages produces solutions that are breathtakingly, devastatingly elegant, beautiful and terse.  When the tool matches the problem, there are very few things in the universe more lovely than a properly designed functional program.  The issue though, is ‘A’ – that the tools are difficult to comprehend, and ‘B’ – I said ‘when the problem matches the tool’, there are lots of problems that don’t match functional programming language tools because functional programming language tools really want to live in a universe where everything is stateless and all transactions are reversible.  Uh oh.  You run into problems with those assumptions the moment you deal with messy things like input/output operations…intrinsically not reversible.  A lot of the complexity in functional programming languages arises from this unavoidable interface, this energy barrier between the programming language’s internal world of pure logic, statelessness and reversibility and infinite backtracking, and the messy exterior world where we actually have to deal with stateful objects…They are fascinating tools, they are good for some classes of problems, I love them aesthetically.  I don’t know if they will ever be more than a small minority preference.”

“Nevertheless, even though these tools may never get huge traction, learn one anyway.  When you get deep enough into any functional language, for me it happened with Lisp…there will come a point at which you will achieve Satori, you will achieve enlightenment about how functional programming actually works and your entire universe will lurch sideways and never be quite the same again.  And even if you never use a functional programming language for a line of code after that, it will change the way you think, it will change the way you form abstractions, it will clean things up.”

“Ironically in the case of Python, Guido actually doesn’t like Lisp and his personal preference would be to take the functional constructs out of the language (Python), but every time he does that a bunch of his friends and senior developers, including me, look at him and say ‘you will take away my lambdas when you pry them from my cold dead fingers’.”

“The advantage of Python over Java is that it’s less heavyweight, there isn’t a huge syntactic thicket of declarations and codicils and stuff that from a point of view of a Python programmer is extraneous junk that gets in the way of comprehension.  There are languages that are much worse that way, see my rant about C++.  Java is a bit too syntax-heavy and cluttered for optimum long-term maintainability.  That is my opinion anyway.”

“We now live in a situation where lots of people can use and enjoy open source tools and an increasing number of people can make a living writing and maintaining them, and that’s a good thing.  Do I want to see all software become non-proprietary?  It wouldn’t particularly bother me if that happened but it’s not a major objective for me.  It doesn’t harm me that other people write proprietary software as long as they don’t try to infringe on my freedom to write software the way I want to.  That’s the freedom I’m concerned with protecting.  I want people who voluntarily choose to be part of the open source community to be able to continue to be a part of it, and as long that objective is achieved, what other people do is not really of much concern to me.”

Your comments are welcomed below, thanks for reading.

One comment

  1. Peter Lund

    ” I have been wondering since about 1971 why nobody took the ball and ran with Hoare’s communicating sequential processes model. ”

    It’s from 1978:

    Lots of people have picked it up. Occam is an early example from 1983. Limbo (a predecessor to Go) is from 1995. Newsqueak (a predecessor to Limbo) is from the late 80’s.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s