[spectre256]

We talked about running a Pelias geocoder AMI back at Mapzen, and it still might make sense today. I'd be really interested to hear how it works out for you.

My personal feeling is that developers are irrationally averse to paying for someone else to set up and reliably run software for them, but that it's usually a massively good deal to pay for that service.

Even a simple software project that takes an hour a month to keep running is well worth most companies paying $100/month to not have to worry about, for example.

[jopsen]

Developers don't want to pay that's true.. I hate a adopting a anything I can't setup myself or get cheap with a pay-as-you-go model.

But as soon as I'm deploying anything in a corporate setting, I would always prefer to pay rather than asking IT to setup and maintain a service. Or worse do it myself..

The combination is good. Because I would rarely try out something I couldn't run myself. But I would also never put it into production.

[eruci]

It has worked out relatively well. (Mostly due to the fact we have no competition on geoparsing )

There are a few geocoding AMI's on AWS Marketplace. ( https://aws.amazon.com/marketplace/search/results?searchTerm... )

Some big companies on the list (Pitney Bowes, Mapbox, Here)

I've always held Mapzen in high regard as one of the best geocoding APIs to come out of the opensource camp. Mapzen would have been a nice addition to AWS for sure.

Some companies prefer having their own server rather than sharing access to an API.

[spectre256]

Oh wow I didn't realize there were so many! And yeah, geoparsing is a huge challenge above and beyond "standard" geocoding. Specialization like that is a great asset.

For us, we've really been focusing on the customization of Pelias for people who have specific needs. We have Elasticsearch under the hood which is perfect for that sort of thing, and the nature of customization is that an AMI probably can't do what our clients need. But probably someday it will make sense.