Nov 21, 2021
Perl was started by Larry Wall in 1987. Unisys had just released the 2200 series and only a few years stopped using the name UNIVAC for any of their mainframes. They merged with Burroughs the year before to form Unisys. The 2200 was a continuation of the 36-bit UNIVAC 1107, which went all the way back to 1962. Wall was one of the 100,000 employees that helped bring in over 10 and a half billion in revenues, making Unisys the second largest computing company in the world at the time.
They merged just in time for the mainframe market to start contracting.
Wall had grown up in LA and Washington and went to grad school at the University of California at Berkeley. He went to the Jet Propulsion Laboratory after Grad School and then landed at System Development Corporation, which had spun out of the SAGE missile air defense system in 1955 and merged into Burroughs in 1986, becoming Unisys Defense Systems.
The Cold War had been good to Burroughs after SDC built the timesharing components of the AN/FSQ-32 and the JOVIAL programming language. But changes were coming. Unix System V had been released in 1983 and by 1986 there was a rivalry with BSD, which had been spun out of UC - Berkeley where Wall went to school. And by then AT&T had built up the Unix System Development Laboratory, so Unix was no longer just a language for academics.
Wall had some complicated text manipulation to program on these new Unix system and as many of us have run into, when we exceed a certain amount of code, awk becomes unwieldy - both from a sheer amount of impossible to read code and from a runtime perspective. Others were running into the same thing and so he got started on a new language he named Practical Extraction And Report Language, or Perl for short. Or maybe it stands for Pathologically Eclectic Rubbish Lister. Only Wall could know.
The rise of personal computers gave way to the rise of newsgroups, and NNTP went to the IETF to become an Internet Draft in RFC 977. People were posting tools to this new medium and Wall posted his little Perl project to comp.sources.unix in 1988, quickly iterating to Perl 2 where he added the languages form of regular expressions. This is when Perl became one of the best programming languages for text processing and regular expressions available at the time.
Another quick iteration came when more and more people were trying to write arbitrary data into objects with the rise of byte-oriented binary streams. This allowed us to not only read data from text streams, terminated by newline characters, but to read and write with any old characters we wanted to. And so the era of socket-based client-server technologies was upon us. And yet, Perl would become even more influential in the next wave of technology as it matured alongside the web.
In the meantime, adoption was increasing and the only real resource to learn Perl was a the manual, or man, page. So Wall worked with Randal Schwartz to write Programming Perl for O’Reilly press in 1991. O’Reilly has always put animals on the front of their books and this one came with a Camel on it. And so it became known as “the pink camel” due to the fact that the art was pink and later the art was blue and so became just “the Camel book”. The book became the primary reference for Perl programmers and by then the web was on the rise. Yet perl was still more of a programming language for text manipulation. And yet most of what we did as programmers at the time was text manipulation.
Linux came around in 1991 as well. Those working on these projects probably had no clue what kind of storm was coming with the web, written in 1990, Linux, written in 1991, php in 1994, and mysql written in 1995. It was an era of new languages to support new ways of programming. But this is about Perl - whose fate is somewhat intertwined.
Perl 4 came in 1993. It was modular, so you could pull in external libraries of code. And so CPAN came along that year as well. It’s a repository of modules written in Perl and then dropped into a location on a file system that was set at the time perl was compiled, like /usr/lib/perl5. CPAN covers far more libraries than just perl, but there are now over a quarter million packages available, with mirrors on every continent except Antartica.
That second edition coincided with the release of Perl 5 and was published in 1996. The changes to the language had slowed down for a bit, but Perl 5 saw the addition of packages, objects, references, and the authors added Tom Christiansen to help with the ever-growing camel book. Perl 5 also brought the extension system we think of today - somewhat based off the module system in Linux. That meant we could load the base perl into memory and call those extensions.
Meanwhile, the web had been on the rise and one aspect of the power of the web was that while there were front-ends that were stateless, cookies had come along to maintain a user state. Given the variety of systems html was able to talk to mod_perl came along in 1996, from Gisle Was and others started working on ways to embed perl into pages.
Ken Coar chaired a working group in 1997 to formalize the concept of the Common Gateway Interface. Here, we’d have a common way to call external programs from web servers. The era of web interactivity was upon us. Pages that were constructed on the fly could call scripts. And much of what was being done was text manipulation.
One of the powerful aspects of Perl was that you didn’t have to compile. It was interpreted and yet dynamic. This meant a source control system could push changes to a site without uploading a new jar - as had to be done with a language like Java. And yet, object-oriented programming is weird in perl. We bless an object and then invoke them with arrow syntax, which is how Perl locates subroutines. That got fixed in Perl 6, but maybe 20 years too late to use a dot notation as is the case in Java and Python.
Perl 5.6 was released in 2000 and the team rewrote the camel book from the ground up in the 3rd edition, adding Jon Orwant to the team. This is also when they began the design process for Perl 6. By then the web was huge and those mod_perl servlets or CGI scripts were, along with PHP and other ways of developing interactive sites, becoming common. And because of CGI, we didn’t have to give the web server daemons access to too many local resources and could swap languages in and out. There are more modern ways now, but nearly every site needed CGI enabled back then.
Perl wasn’t just used in web programming. I’ve piped a lot of shell scripts out to perl over the years and used perl to do complicated regular expressions. Linux, Mac OS X, and other variants that followed Unix System V supported using perl in scripting and as an interpreter for stand-alone scripts. But I do that less and less these days as well.
The rapid rise of the web mean that a lot of languages slowed in their development. There was too much going on, too much code being developed, too few developers to work on the open source or open standards for a project like Perl. Or is it that Python came along and represented a different approach with modules in python created to do much of what Perl had done before?
Perl saw small slow changes. Python moved much more quickly. More modules came faster, and object-oriented programming techniques hadn’t been retrofitted into the language. As the 2010s came to a close, machine learning was on the rise and many more modules were being developed for Python than for Perl.
Either way, the fourth edition of the Camel Book came in 2012, when Unicode and multi-threading was added to Perl. Now with Brian Foy as a co-author. And yet, Perl 6 sat in a “it’s coming so soon” or “it’s right around the corner” or “it’s imminent” for over a decade.
Then 2019 saw Perl 6 finally released. It was renamed to Raku - given how big a change was involved. They’d opened up requests for comments all the way back in 2000. The aim was to remove what they considered historical warts, that the rest of us might call technical debt. Rather than a camel, they gave it a mascot called Camelia, the Raku Bug.
Thing is, Perl had a solid 10% market share for languages around 20 years ago. It was a niche langue maybe, but that popularity has slowly fizzled out and appears to be on a short resurgence with the introduction of 6 - but one that might just be temporary.
One aspect I’ve always loved about programming is the second we’re done with anything, we think of it as technical debt. Maybe the language or server matures. Maybe the business logic matures. Maybe it’s just our own skills. This means we’re always rebuilding little pieces of our code - constantly refining as we go. If we’re looking at Perl 6 today we have to look at whether we want to try and do something in Python 3 or another language - or try and just update Perl. If Perl isn’t being used in very many micro-services then given the compliance requirements to use each tool in our stack, it becomes somewhat costly to think of improving our craft with Perl rather than looking to use possibly more expensive solutions at runtime, but less expensive to maintain.
I hope Perl 6 grows and thrives and is everything we wanted it to be back in the early 2000s. It helped so much in an era and we owe the team that built it and all those modules so much. I’ll certainly be watching adoption with fingers crossed that it doesn’t fade away. Especially since I still have a few perl-based lamda functions out there that I’d have to rewrite. And I’d like to keep using Perl for them!