I am a network administrator. I work on a team of fewer than a dozen people, only one of whom is a full-time developer (and it’s not me). But I do write code — software that supports the operational needs of my organization; software that helps me do my job. That code has a lifetime of years; some of it is already in its second decade. I need stable programming environments. I need to be able to write something, get it working, and then not have to think about it until my environment changes. I do not have a staff of fifty. I do not want to do what the “cool kids” are doing; I am an administrator, and what I do will never be “cool”. But the “cool kids” will not be able to do anything if my infrastructure doesn’t work right, consistently, efficiently, over the long term. I do not have $2 million in VC money to throw down a rathole while you rototill some language, package, or operating-system component I depend on a dozen times a year. Cool new stuff is nice, everybody wants cool new stuff once in a while. So long as the old stuff (which was cool new stuff itself once) doesn’t break. So long as it happens on a predictable schedule — measured in years or quarters, not months, not weeks, not days.

The people I support are going to work on one thing, and produce one deliverable (a Ph.D. thesis), over the course of three to seven years. They (or their advisors) would like to spend their money on inventing their own cool new things. They get unhappy when it gets spent on staff time to do unproductive things like rewriting infrastructure software to catch up with development environments that assume anyone who writes a line of code works on a team of fifty full-time developers. My colleagues and I don’t. We are sysadmins, and network admins, and we keep the bits flowing and empty the bit bucket when it gets full. On occasion, we’ll build infrastructure that enables someone to do something really important and groundbreaking. There are more of us than there are hot new Silicon Valley Web startups, and we’re tired of being taken for granted by developers.

Thank you for your attention.