This week I finished off a quick project to upgrade the University's instance of CQPweb, a web app for doing corpus linguistics - the analysis of language based on large corpora of texts from sources such as journalism, legal proceedings, TV scripts and the like. (One of the first things you learn when you work with linguists is that the plural of "corpus" is "corpora".)

CQPweb is a web front-end to a command-line tool whose basic job is generating concordances - finding all occurences of a word in a corpus and returning the results with some context. It's very web 1.0 - there's hardly any JavaScript, just an SQL database and a lot of PHP to marshall query results into HTML. It's the sort of application which people can be dismissive about - in an unfashionable language, not doing anything fancy with machine learning or natural language processing. But keeping it going is work worth doing: a better user of time than trying to reimplement it in a modern stack, which would just mean needing to get rid of all the bugs again.

There's a fair bit of nostalgia around these days for the old, small web, and putting a piece of it into Docker and dealing with all of the odd edge cases which this raised bought up some of that nostalgia, along with a bit of career vertigo. Looking after - not to mention writing - this sort of thing was what I did in my first job at a university, when we were all lot more relaxed about security and caring whether we could reproduce an environment.

Hopefully, this means that the next upgrade will take less effort.