Following up on my finger-wag at Google for not properly supporting mashup developers by messing up versioning, I have to now send them full props for one thing they’re doing very, very, right.

One half of Postel’s Law says “Be liberal in what you accept”, and Google has done exactly that in at least two places. First is in Google Maps, where you can enter pretty much anything resembling a street address, and more often than not it’ll grok it. That’s not to suggest it couldn’t be improved mind you – about a quarter of the time I probably have to refine what I enter, but still, that’s not bad. Without this capability, Maps mashups would be a lot more difficult to develop in part because there exists no widely adopted standard format for an address, leaving prose as the only option for interchange. By doing this Google is absorbing the costs of solving the problem, and relieving mashup developers of the burden. Quite the contrast to their API versioning policy! 8-O

Another example of this I’ve noticed is Google Calendar, where it can accept dates also in prose, even relative ones like “tomorrow”. And this is despite having somewhat decent time and calendaring standards. So why the prose? It just simplifies integration, as the calendar integration with GMail demonstrates; it can pick out dates from an email without requiring the sender conform to any particular standard. Actually, I don’t know if that’s GMail or calendar doing it, but I hope it would be the calendar so that it can be more easily reused in other calendar-integration scenarios.

FWIW, I recall Peter Norvig saying something in his recent highly publicized run-in with TimBL about the value of this approach (mining existing content) over authoring new content; just can’t find the quote I’m looking for right now, but I’ll add it when I do.

I previously pointed to the announcement of the shutdown of a version of the Google Adwords API, and commented that this really isn’t the way to go about versioning your Web 2.0 APIs.

I’ve been digging into Google Maps recently, and noticed that they’re making the same mistake;

The v=2 part of the URL http://maps.google.com/maps?file=api&v=2 refers to “Version 2 ” of the API. When we do a significant update to the API in the future, we will change the version number and post a notice on Google Code and the Maps API discussion group.

After a new version is released, we will try to run the old and new versions concurrently for about a month. After a month, the old version will be turned off, and code that uses the old version will no longer work.

Obviously we’re still pretty early into the whole “mashup” thing, but not too early that we shouldn’t be thinking about best practices IMO. And best practice #1? Don’t do what Google’s doing here, which is asking all their users – mashup developers who have committed themselves to this service – to absorb the cost of their inability to develop an extensible API.

I think a good rule of thumb for service providers is to assume that you’ve got a million mashups using your service, and therefore that the cost of incompatible changes is prohibitive for the mashup developers. Any other approach is sure to drive those developers to other service providers who do a better job evolving their APIs; bad juju if your business depends on attracting eye-balls.

So what constitutes a “better job”? I suggest it has something to do with declarative Javascript, but my argument needs some work so I’ll save that for another post.

I’m officially unemployed now. I’m not quite sure what I’ll be doing next, but I’ve already been contacted about doing some consulting work in the enterprise Web services space. Consulting would be “ok”, but it would be nice to have the relative comfort of a “9 to 5” job for a little while after a year of doing my own thing. Plus I think I would have a bigger impact if I was with an established company with customers to satisfy.

So if you’re interested in the services of a large scale distributed systems expert who knows why Web services will fail, and how to fix them, drop me a line.