Way back when I was first introduced to application development, I remember seeing “code completion” for the first time and thinking that it was the single coolest thing I’d ever seen. Now setting aside the fact that that said more about the technology I’d seen at that point than it did about its importance, I to this day remain mystified why code completion, autocomplete, call it what you will, has not become common enough to deliver as an operating system service.
Part of it I’m sure is the difficulty of scope. For example, text prediction functionality in mobile phones is to me more of a hindrance than a help, simply because it has no context for what I’m typing, nor can I improve its context via customization. As a result, autocomplete is typically surfaced to us today via two different approaches designed to limit that scope: completion off of a known, standard vocabulary (e.g. Gmail’s contact address autocompletion or reserved words in an IDE) or off a user authored and maintained dataset (e.g. Typeit4me).
What’s still surprising to me, however, is that given how much of a boon autocomplete can be, that we have yet to see widespread interest in the collaboration/information worker/etc space embracing the technology (yes, I’m aware that Open Office does this to some degree). While it’s great to see individual implementations for specific services (here’s one for del.icio.us), I’d much prefer to have some centralized control for autocomplete more broadly. It would be nice, for example, to have an autocomplete that worked for me across MT, Flickr, and del.icio.us. I’m sick of typing “