Mon, 09 Feb 09

Idea - Web Services Could Publish Stats About the Clients That Connect to Them

It’s not always obvious whether a client library is still compatible with the web service it was developed for. I’ve seen this before with flickr libraries and came across it again a couple of weeks ago when playing around with Amazon Simple DB. I found three (although there are more) ruby clients that supposedly connected to the Amazon Simple Db service: NY Times, AWS SDB and Right AWS. Unfortunately, without trying them (which, granted, wouldn’t have taken too long) I had no idea whether they were going to work or do what I needed. The fact that none of them appeared to support the latest version of the request signing protocol made me question how useful they’d be. It struck me that it should be possible for the web service to know which client is being used to connect assuming the client supplies some sensible identifiable information (in the user agent http header for example). If the provider was to publish this information (cient name/version, total requests, date of last request and probably some other useful info) then it should be possible to see which libraries still work with the web service. Does anyone do this? Does it actually sound like a sensible idea?