Why bother with Scripting?

Many years back, Larry Wall shared his thesis on the nature of scripting. Since recently even Java gained 'script' support I thought it would be fitting to revisit the topic, and hopefully relevant to the perl and raku language community.

The weakness of Larry's treatment (which, to be fair to the author, I think is more intended to be enlightening than to be complete) is the contrast of scripting with programming. This contrast does not permit a clear separation because scripts are programs. That is to say, no matter how long or short, scripts are written commands for a machine to execute, and I think that's a pretty decent definition of a program in general.

A more useful contrast - and, I think, the intended one - is between scripts and other sorts of programs, because that allows us to compare scripting (writing scripts) with 'programming' (writing non-script programs). And to do that we need to know what other sorts of programs there are.

The short version of that answer is - systems and applications, and a bunch of other things that aren't really relevant to the working programmer, like (embedded) control algorithms, spreadsheets and database queries. (The definition I provided above is very broad, by design, because I don't want to get stuck on boundary questions). Most programmers write applications, some write systems, virtually all write scripts once in a while, though plenty of people who aren't professional programmers also write scripts.

I think the defining features of applications and systems are, respectively:

  • Applications present models to users (for manipulation)
  • Systems provide functionality to other programs

Consider for instance a mail client (like thunderbird) in comparison to a mailer daemon (like sendmail) - one provides an interface to read and write e-mails (the model) and the other provides functionality to send that e-mail to other servers.

Note that under this (again, broad) definition, libraries are also system software, which makes sense, considering that their users are developers (just as for, say, PostgreSQL) who care about things like performance, reliability, and correctness. Incidentally, libraries as well as 'typical' system software (such as database engines and operating system kernels) tend to be written in languages like C and C++ for much the same reasons.

What then, are the differences between scripts, applications, and systems? I think the following is a good list:

  • Scripts tend to be short, applications in particular can grow very large.
  • Scripts tend to be ad-hoc (written for a specific need), applications and systems tend to be designed for a range of use cases. (Very common example: build scripts)
  • Scripts tend to run only in a specific environment; in contrast, many applications are designed for a range of devices/clients; many systems have specific requirements but the intention is that they can be setup in multiple distinct environments.
  • Because scripts are ad-hoc, short, and environment-dependent, many of software engineering standard best practices don't really apply (and are in fact often disregarded).

Obviously these distinctions aren't really binary - 'short' versus 'long', 'ad-hoc' versus 'general purpose'  - and can't be used to conclusively settle the question whether something is a script or an application. (If, indeed, that question ever comes up). More important is that for the 10 or so scripts I've written over the past year - some professionally, some not - all or most of these properties held, and I'd be surprised if the same isn't true for most readers. 

And - finally coming at the point that I'm trying to make today - these features point to a specific niche of programs more than to a specific technology (or set of technologies). To be exact, scripts are (mostly) short, custom programs to automate ad-hoc tasks, tasks that are either to specific or too small to develop and distribute another program for.

This has further implications on the preferred features of a scripting language (taken to mean, a language designed to enable the development of scripts). In particular:

  • It should make programs concise. The economic rationalization is that the total expected lifetime value of a script, being ad-hoc and context-dependent, is not very great, so writing it should be cheap, which implies that the script should be short).
  • Related to this, the value provided by type systems is generally less than in larger (application) programs, and the value of extensive modelling features (class hierarchies) is similarly low, so many scripting languages have very weak type systems and data modelling features, if they have them at all.
  • Interoperation with the environment is on the other hand very important, so I/O features tend to be well-developed. (Contrast C, in which I/O is entirely an afterthought provided by a library).
  • It is acceptable to depend on a local environment in implicit ways, since that's what you are going to do anyway.
  • It is acceptable to warn on a condition that might've been a fatal error in another programming language.
  • In fact, I think that concerns of correctness are often different, meaning relaxed, compared to applications, again because scripters don't necessarily expect their scripts to run on every environment and with every possible input.
As an example of the last point - Python 3 requires users to be exact about the encoding of their input, causing all sorts of trouble for unsuspecting scripters when they accidentally try to read ISO-8551 data as UTF-8, or vice versa. Python 2 did not, and for most scripts - not applications - I actually think that is the right choice.

This niche doesn't always exist. In computing environments where everything of interest is adequately captured by an application, or which lacks the ability to effectively automate ad-hoc tasks (I'm thinking in particular of Windows before PowerShell), the practice of scripting tends to not develop. Similarily, in a modern 'cloud' environment, where system setup is controlled by a state machine hosted by another organization, scripting doesn't really have much of a future.

To put it another way, scripting only thrives in an environment that has a lot of 'scriptable' tasks; meaning tasks for which there isn't already a pre-made solution available, environments that have powerful facilities available for a script to access, and whose users are empowered to automate those tasks. Such qualities are common on Unix/Linux 'workstations' but rather less so on smartphones and (as noted before) cloud computing environments.

Truth be told I'm a little worried about that development. I could point to, and expound on, the development and popularity of languages like go and rust, which aren't exactly scripting languages, or the replacement of Javascript with TypeScript, to make the point further, but I don't think that's necessary. At the same time I could point to the development of data science as a discipline to demonstrate that scripting is alive and well (and indeed perhaps more economically relevant than before).

What should be the conclusion for perl 5/7 and raku? I'm not quite sure, mostly because I'm not quite sure whether the broader perl/raku community would prefer their sister languages to be scripting or application languages. (As implied above, I think the Python community chose that they wanted Python 3 to be an application language, and this was not without consequences to their users). 

Raku adds a number of features common to application languages (I'm thinking of it's powerful type system in particular), continuing a trend that perl 5 arguably pioneered. This is indeed a very powerful strategy - a language can be introduced for scripts and some of those scripts are then extended into applications (or even systems), thereby ensuring its continued usage. But for it to work, a new perl family language must be introduced on its scripting merits, and there must be a plentiful supply of scriptable tasks to automate, some of which - or a combination of which - grow into an application.

For myself, I would like to see scripting have a bright future. Not just because scripting is the most accessible form of programming, but also because an environment that permits, even requires scripting, is one were not all interesting problems have been solved, one where it's users ask it to do tasks so diverse that there isn't an app for that, yet. One where the true potential of the wonderful devices that surround is can be explored.

In such a world there might well be a bright future for scripting.

Reacties

Populaire posts van deze blog

New years post

Reverse Linear Scan Allocation is probably a good idea