albarrentine
commented
6 years ago On a scale of "0: nothing has changed" to "10: everything is ruined" it's about a 2 or a 3. Libpostal was more loosely affiliated with Mapzen than some of their in-house projects that had full-time people working on them. I was never a Mapzen employee, though they did provide the AWS resources for this project and funded certain aspects of its development (like our recent deduping release, which got out the door before the shutdown). I'm personally using libpostal in my work on voting rights, so it will continue to be developed.
That said, producing something as data-intensive as libpostal is certainly not free, so for companies using libpostal and benefitting from it, there's never been a better time to make a monthly contribution to our OpenCollective: https://opencollective.com/libpostal.
At present I'm cataloguing the baseline expenses and looking at ways to reduce the costs of the data pipeline behind libpostal and storing the models/data that people download during install, so that I'm not picking up the check for everyone as it were. Part of that will include moving the model releases to Github (which allows files up to 2GB as part of the release) and the training data to the Internet Archive. Have also started to re-architect the training data generation using ElasticMapReduce so it won't require any long-running machines. However, that still costs money, and once the training data is built, model training currently takes around 7 days, which will also require some compute time. All of this will be coming out of the OpenCollective budget now, so I'm hopeful our users will step up and contribute to the project's sustainability post-Mapzen.
jamiehutton
Hi Al,
I got an email today last week about Mapzen shutting down - i just wanted to check whether the openvenues/libpostal project is impacted in any way by this?
Jamie