SIGUSR2 home apg(7) colophon feed

% Fabric not Uniforms % advice, rant % 2012-10-11

Too many decisions in the software industry are made, not by rigorous evaluation or personal experience, but instead by fads that a small, vocal minority tout.

These things aren’t typically as simple as the best sorting algorithm, or a recipe for hashing passwords1 (maybe a bad example[^2]), but are deeper than that–I’m talking about which programming language one should use, or which “framework,” or even which message broker or database.

Many times the question about whether the problem’s solution even requires a framework, or a message broker isn’t discussed. “Of course we need X, GitHub uses it!” Or “Node.js achieves 3,000 requests/s in NodeLover’s (pointless) benchmarks! Surely it’s the right tool for our factorial server!”

Please, make the herding stop!

Ian MacKaye, front man for bands like Minor Threat, The Teen Idles and Fugazi, coined the term “straight edge” in a Minor Threat song he wrote called, well, “Straight Edge.” In that biography, he affirms that he “laughs at the thought of eating ludes” and he has “better things to do, than sit around and fuck my head.” With “Straight Edge” and another song “Out of Step”, in which MacKaye affirms “I don’t smoke. I don’t drink. I don’t fuck. At least I can fucking think” MacKaye inadvertently started a movement, or a life-style, or a music scene–depending on whom you talk to[^3].

He often speaks about how those two songs were re-purposed to power other people’s agendas, which he never intended. There’s a great metaphor that goes with it:

…if words or songs or lyrics are clothes, then really direct ideas become uniforms that anybody can put on. So some people can use the uniform to further their agenda and they don’t engage with it, they just put it on. And in my mind I thought being really direct would make it much more difficult for people to abuse the idea, but actually it’s easier to abuse because the ideas were finished and they didn’t have to contribute any of their own selves to the situation. -[Ian MacKaye][4]

In the software world there are “uniforms” too. Many of us choose to put on the Rails uniform, the NoSQL uniform, the Node.js uniform, or even the Agile uniforms. These tools and methodologies have movements of course, with people actively recruiting for them at conferences, via blog posts, books and more. Like fish, we too are attracted to shiny buttons.

Fabric, on the other hand is a foundational idea. It’s non-biased research, from industry, from academia, from individuals on blogs. It’s the type of talk at a conference which discusses concurrency primitives in a multitude of languages, their pros, their cons. It’s a paper about a new data structure which outlines it’s performance properties under many different circumstances.

Uniforms hinder progress in our industry. The legions that uniforms create can’t see past their solutions–sort of analogous to “the blub paradox.”[^4].

Uniforms don’t always fit[^5], and more often than not cause us to lose site of the original problem. Instead we focus on how we can use a belt and some thread to make the uniform fit our new problem, for no better reason than allegiance. This leads to security problems, bugs and bloated software, among other things. People too often tread water dressed as a Policeman.

We need to all become better tailors. And we certainly need to learn the art of fabric construction and selection. Only then will we be able to create our own properly fitting clothes.


  1. “Just use bcrypt” – On Cryptography and Dogmas, Old, but epic fad herd

    [^2]: A potential bad example because even if you don’t understand it, using bcrypt is fundamentally more secure than salted, or even non-salted SHA1 / MD5 for storing passwords. But the whole “don’t think or evaluate for yourself, ‘cause we’re waving what we use at you and we’re awesome” attitude is shit.Other obvious choices are scrypt, PBKDF1 or PBKDF2.

    [^3]: EDGE Perspectives on Drug Free Culture has great explorations on the matter

    [^4]: Maybe it’s a stretch, but the comparison blub seems apt.

    [^5]: But, there’s no denying that a uniform sometimes perfectly fits. The reason for this stems from the fact that uniforms are normally patterned after a successful project. Rails was patterned after Basecamp was created, for instance. Django was patterned after a CMS for a newspaper. There are many cases when Rails and Django (and other frameworks and languages) are absolutely the best choice for a project.

    [1]: #bcrypt

    [2]: #security

    [3]: #edge

    [4]: http://gothamist.com/2006/11/06/interview_with.php

    [5]: #blub

    [6]: #fit

    [7]: http://antirez.com/post/crypto-dogmas.html

    [8]: https://github.com/antirez/lamernews/pull/8

    [11]: http://basecamp.com