If you put a million monkeys at a million keyboards, one of them will eventually write a Java program. The rest of them will write Perl programs.
— anonymous Internet wag
I was a laser printer guru in 1998. I spent most of my time talking to customers about printing and imaging. I can probably find a blocked paper sensor on an HP LaserJet 5si ten years after my last experience troubleshooting the paper jam message. (I can probably close my eyes and walk someone through it from memory.)
I delivered two programs in that job. One was for the customer service department. Every time someone from level one support had to ask a question of level two support, level two support needed to record the relevant product line. These became daily, weekly, and monthly reports used to bill the subdepartment for each product for the service calls.
I was teaching myself Java at the time. I wrote a small AWT GUI application (in these days, that's what you could get with the Java GNU/Linux port on Red Hat 5.3) which ran on a spare PC we set up at the level two support desk. It logged all statistics to a flat file which another program -- I believe a Tcl program written by a co-worker -- summarized into the reports. This program took a couple of weeks to write. The program was still running when I last walked through that floor of that building in early 2000.
My second program was for the networking support group. In those days, internal support groups often managed the external-facing web site for their products. They wanted a program which could identify when a page or pages in the web site changed and email an opt-in list of customers.
I thought about the problem and wrote a small Bourne shell script (running on HP-UX 10.x, I believe) to do the job. The code must have been between ten and fifteen lines long. As it happens, the networking group used IIS and an early Java application server plugin to manage their web site, so they wanted a Java program which ran on Windows. They asked me to port my proof of concept to Java instead.
I never finished that program. I switched to a system administrator role and discovered that the Perl I'd dabbled with late in 1998 was a much better fit for the system administration and web development I needed to do.
The afternoon of my first day on the new job, the administrator I was replacing turned around and asked me if I knew Perl. I'd dabbled. I'd written a bit. I could read a program and have some sense of what was happening. I said "Some, yeah."
He was working on a program and kept getting a weird syntax error. I looked over his shoulder and said "That
elseif should be an
As I went through the files he left, I found several printouts of Perl programs he'd downloaded from the Internet, changed the copyrights, and tweaked very slightly.
If the department I worked in back then still exists (and it might) and if someone still remembers me from then (doubtful), I'd have no surprise to learn that some of the code I wrote there still exists. (One evening I saved my group a million dollars in penalties by fixing a misbehaving FTP server right before a deadline to deliver new firmware images to a vendor. They gave me a plaque.)
I've toyed with Java some since then, but I haven't delivered any significant software in Java. I hear it's a passable language with some great libraries and effective tools.
I've spent the intervening years understanding software development, helping create a testing culture, figuring out what's wrong with object orientation and how to fix it, and learning how virtual machines work to help invent the future (or at least drag mainstream dynamic languages kicking and screaming into the '90s).
I delivered that first Java program because I was stubborn and had a lot of spare time over that fortnight.
I've delivered almost every Perl program since then because of Perl's whipupitude factor. That's how the administrator I replaced could download a likely program, tweak a few things, and pass off a working solution as his own. That's how a motivated, mumble-something fledgling music performance major like me could write a proof of concept mail notification service in an hour with the Bourne shell and port it to Perl 5 in an afternoon and still not be sure which Java SMTP package actually worked back in 1998.
I have no illusion that I wrote idiomatic Perl 5 code then. (The Perl Renaissance had yet to begin.) I doubt I'd want to maintain that code now. I wouldn't suggest it as a good example for novice developers to emulate.
Yet it did the job. The sound of that code is the sound of me and my users and co-workers not banging their heads against the wall. It's the sound of a 30 second task repeated ten times a day -- and its concomitant context switch -- vanishing.
Anonymous Internet critics can shake their heads and wag their beards and tear their outer garments and throw dust in the air as they wail and weep that ugly Perl code exists and persists.
I'm all for improving Perl code. I believe in tests and refactoring and idioms and reuse and clarity and good naming and succinctness and encapsulation and proper factorization and favoring simplicity over cleverness. I believe in teaching novices and guiding them and encouraging them to develop good habits and consider maintenance and understand what happens and how things work and to think not only about what they want to do but why and how to do so sustainably.
I want better defaults and clearer error messages and fewer rough patches where they can trip and stumble and skin their knees and elbows.
I want that very much.
Yet the whipupitude that lets novices -- and there are some six and a half billion people on the planet who've never even thought they could write a useful computer program -- do useful things even if sometimes they do it the ugly way is a positive aspect of Perl. It's a positive aspect of many tools.
By all means let's continue making Perl easier to understand and gentler to learn and more encouraging to write good code the right way. That's important. We're all novices sometimes. Let's help them get their work done and then help them get their jobs done well. Programming well is difficult; why not encourage them with small victories? Why make the barrier to accomplishment artificially high?
I don't understand that argument.
Then again, some people probably think that Pablo Neruda never existed after hearing me speak Spanish.