Theme, Pragmatics, and Purpose in Programming Language Implementation
I first wrote this in 2011, revised it in 2015, and rewrote it again in 2025 and 2026. The argument has not changed, but the evidence is clearer.
A programming language needs a reason to exist. Not a slogan, but a purpose specific enough that a designer can use it to say no to a feature, and specific enough that a programmer can use it to decide whether to learn the language at all.
Where This Argument Comes From
A thread runs through Masterminds of Programming, the book of interviews with language designers reviewed on this page: designing a language for change and extension is difficult. A designer must evaluate many branching paths and risk many ways to go wrong. The taste required to navigate this plan is also one of the strongest elements of a language that lasts.
Purpose is how you choose among those paths. A language that knows what it is for can decide what it needs to be extensible toward.
Let's try an analogy.
Superheroes and Purpose
The first season of Heroes had a motto: "Save the cheerleader, save the world." It was direct, and it made the stakes both human (protect one young woman) and enormous (avert a catastrophe). Later seasons lost that clarity and lost their audience.
A realistic hero needs an obvious motivation. Without the desire to use his power responsibly, and the memory of his Uncle Ben, Spider-Man is just another selfish teenager with superpowers. Superman is nearly indestructible, so the stories that work are about why he is Clark Kent and what he owes the world he landed in.
Take away the reason for the superhero and the stakes of the stories, and the resulting plots jerk the character around for the sake of plot.
Software and Purpose
Any software project without users, without a specific reason to exist, is in danger of becoming a perpetual research project.
There is nothing wrong with programming language research. The world needs more of it, because it improves the languages that real programmers use to build software for real users. But a language is interesting in practice only when you can do things with it.
JavaScript would be a dead end as an extension language except for the fact that every browser ships it. PHP would be a has-been templating system except that it is still among the simplest ways to put dynamic pages on a server. JavaScript and PHP didn't win on spectacular technical merits of language design and implementation. They won because they were good enough at specific jobs.
What Happened to Perl 6
Larry Wall announced the design process for Perl 6 in July 2000. The first stable release of the language specification came on Christmas Day 2015. In October 2019 the project renamed itself Raku, to separate it from Perl 5, which it had never replaced. (Sources: Wikipedia's Raku article, which cites contemporary coverage of the rename.)
When I first wrote this article, the question was why Perl 6 existed. "It should be better than Perl 5" was an incomplete answer: better how, measured along which axis, for whom? Today the question has an unflattering answer.
It is 2026. Raku has achieved no significant, measurable adoption. It does not appear in the top 100 of the TIOBE index (October 2026). The 2025 Stack Overflow Developer Survey does not list it. In my view it has not met the performance goals promised for it over the years. I know of no authoritative benchmark showing that it has, and the one public release-over-release benchmark I found (a single script, measured across Rakudo releases from 2016 to 2022) shows big speedups through 2020, then a sharp regression in late 2021 that had not recovered by the end of 2022.
It may be dangerous to assume that any subsequent improvements would have been trumpeted to the world. The fact of their absence is evidence of a sort.
Raku also remains tied to Perl. It sits under The Perl and Raku Foundation, which holds legal responsibility for both languages, and it shares the Perl and Raku Conference. The rename was meant to give Raku a separate identity. In practice the institutional link is still tight. Raku may do something interesting in the future, but it has been a decade and a half since the first version of this article, and the language remains barely relevant.
Its strongest influence seems to be as an example of how not to run a language project or a migration. I do not know whether its design and especially its virtual machine were the dead ends that some predicted. That is still unresolved. Yet the other prediction came true: a long, slow, and expensive development cycle that would achieve little adoption.
For the project's own account, see the Raku FAQ. It describes Raku as the definition of the language and Rakudo as an implementation of it, and it calls Perl and Raku "sister languages" that share much of their scope but interoperate rather than one replacing the other.
Purpose Done Well
I have not worked on the following projects and I do not claim authority over their design. These are observations from outside.
- Go is well designed for network services deployed in hardened, slimmed, distroless containers, despite a genesis before Docker and Kubernetes. That is a specific purpose, and its restrictions follow from it.
- Swift took over iOS development from zero to dominant in a short time. It had a platform that needed it and an owner willing to move the whole platform.
- Mojo, introduced by Chris Lattner's company Modular in 2023, reimagines the implementation of a Python-like language in service of a purpose, running AI workloads at systems speed. Whether it succeeds is not yet known, but it demonstrates the value of rethinking the implementation to serve a purpose.
Each can finish the sentence "this language is for ____" in a handful of words. Perl 6 never could.
Politics and Pragmatics of Free Software
Free software exists to be used: to enable the freedom of users by replacing proprietary software, or to forestall the popularity of competing proprietary software. Without utility there is neither political nor practical point.
Free software is and is not subject to market pressure. Users will adopt it if it is pragmatically advantageous. Developers, however, may keep developing regardless of any practical use, because volunteers maintain a project for as long as they stay interested. That freedom is valuable, because free software serves as the research arm of the software industry. It is also how a project can spend a decade and a half looking for a purpose and a decade afterward struggling to find relevance.
Call an experiment an experiment at the start, or have the courage to reclassify it as one when the pragmatics are still far off. Research does not and should not have to follow a project schedule. A replacement, however, needs its purpose established before it launches, or you kill off the old version and drive away its users.
Imagine a world where Perl 6 had been a long-lived research project which
fed its results back into Perl 5 development. Obviously some ideas are trivial
to backport (say), some ill-advised (smart match), and others need
continual refinement (classes and roles), but avoiding the perception
of Perl's obsolescence by not declaring Perl 5 obsolete and then not shipping a
viable replacement for at least a generation did not work.
Artistic Expression Is a Luxury
Elegance in a programming language is a luxury you get after you satisfy your
pragmatic constraints. No one cares that Ruby feels elegant if you cannot use
it to script a page in a browser. No one cares that JavaScript steals a little
of your soul every time you forget what this is bound to, because
it is the only option you have. The real question is rarely which tool is the
singular right one. It is which of many tools is good enough for the job.
"We can't tell you what it's for, but you'll know when you need it" is an apology for lacking a design principle.
What Success Means in the LLM Era
Success in the mid-2020s has become a different problem. Network effects are still essential, but people now lean on AI tooling to write code. That changes what wins. PHP is still everywhere because deploying it is simple: W3Techs reports it on about 69% of the sites whose server-side language it can identify (November 2026). JavaScript is growing faster because of its network effects: 66% of respondents to the 2025 Stack Overflow Developer Survey use it. Python gained seven percentage points in that survey, continuing a resurgence that began around 2016 and 2017 with machine learning and, later, LLM tooling.
Easy onboarding, easy deployment, good libraries, and automation win. Those are all pragmatic qualities. Pragmatics and language design in the post-LLM era is a question for a different article.