I’ve previously called in-betweeners pipes (as in Unix pipes) or transformers, but I wanted to stop doing that because I didn’t want to focus on the chaining. Instead I wanted to think about drop-in technology encapsulated into web services.

Some examples are:

• scanR, the top two images. It takes a messy image of, say, a whiteboard, and image processes it to straighten it, crop it, and make the writing clearer.
• Lazymask offers a simple service: Creating paths to chop away backgrounds in Photoshop is tedious work, and expensive. Lazymask let you outsource it cheaply. Send off an image and get it back with the paths drawn in.
• Scanr filters are in-betweeners that transform your MMS images to clean them or perform weird changes.

Okay, I have two comments about these in-betweeners. First, they require almost no interface. In fact, they shouldn’t require any interface. I should be able to drop them into other applications with ease. How? Perhaps whenever I tag a photo with ‘whiteboard’ at Flickr, it’s automatically piped through scanr and the returned image added to my stream. Perhaps if I tag it ‘whiteboard+ocr’ then it’s piped through scanr and then turned into machine-readable text. Likewise if I tag ‘lazymask’ or whatever. I’m not sure how this would work. We’d need some standard call-response protocol, but that’s just convention. It can happen anytime.

Second, I don’t yet understand the business model. It’s like the Amazon S3 thing: how can my website look after billing a user on behalf of S3 or Scanr? What secure and easy protocols could we use for that? I’d love to make a stack of in-betweeners that other people could hook into their applications – or rather, that users could hook into the applications themselves in a Greasemonkey-for-in-betweeners kind of way – but I can’t yet see how it’d work.

Everything I’ve mentioned so far has been image based. There are text in-betweeners too.

Sam Ruby’s Venus filters sit at the head of a stream of Atom/RSS entries and pipe them off… this one to the full-text search, this other to the reblog, these ones to the ego-feed. I’ve talked about weblogs as a series of pipes and how this can be implemented in Atom before, so I’m pleased to see these parts broken out.

There are also MP3-to-text transcription services and the possibilities of data-mining yourself… my email is already piped through spam filters to change it and update my spam database; what else could my email be piped through?

I look forward to a day where interface-less web apps are just as important as interface-ful ones.

Matt Webb, S&W, posted 2006-09-21 (talks on 2006-09-03, 2006-09-17)