Howard Chu @ SymasRE: https://mastodon.social/@nixCraft/116188939207308679
Happiness is watching the 800lb gorilla in your industry careen like a drunkard from one critical mistake to another, over and over
Csillag TamásOracle never really affected us in OpenLDAP much, until they bought Sleepycat and took over BerkeleyDB in 2006. For a couple years the Sleepycat team still ran things there but I suppose eventually their golden handcuffs came off. In 2008 Oracle started sending us threatening emails about needing a license to use BerkeleyDB in OpenLDAP. I forwarded a copy to Keith Bostic asking him wtf and he assured me it was a mistake; as an open source project we didn't need to sign a license with Oracle.
The matter seemed to be resolved but the following year Oracle sent another threatening email. So the desire to move off BerkeleyDB was firmly planted by 2008... Around that time we got contacted by MySQL, to do a joint project to build an OpenLDAP backend on their new NDB Cluster. This seemed pretty promising as a solution for scaling OpenLDAP capacity and soon we had a usable back-ndb. Then Sun bought MySQL, and the project grew to a partnership with Sun/OpenDS too. https://lists.openldap.org/hyperkitty/list/[email protected]/thread/RS7KHPQLNCDYB35K2EJUWNPQBNLQRCJD/?sort=thread
Then Oracle bought Sun and axed the project. The rug was pulled out from under us, and back-ndb was killed.
https://en.wikipedia.org/wiki/Acquisition_of_Sun_Microsystems_by_Oracle_Corporation
So by 2010 we were pretty sick of Oracle messing up our plans. Developing #LMDB wasn't just a technology experiment, it was a strategic step in getting away from Oracle's influence.
You can well imagine how satisfying it's been to see, some years later, that Oracle has deprecated their own use of BerkeleyDB in favor of #LMDB https://mastodon.social/@hyc/115984080628049520
Of course Oracle kind of did it to themselves too, when they changed the BDB license to AGPLv3 in 2013. This prompted Debian to look for alternatives, and #LMDB emerged as the only suitable candidate.
https://lwn.net/Articles/558154/
A bonus from modeling LMDB on the BDB API - we did this to ease development of back-mdb based on back-bdb. But that also meant it was easy for every other project using BDB to migrate too. And after these licensing games, they were eager to migrate so LMDB use exploded.
Bill Fellows@hyc Ha! I had no idea of the convoluted history with the project. Not at all surprised given how I have yet to hear anyone ever say "I had a really good relationship with Oracle. Their pricing was fair and transparent. And they really respected us as a customer/partner"
@billinkc lol indeed, words that no one has said, ever
Jan Lehnardt 
@hyc finally you can send them a threatening letter
JP Mens@jpmens Larry Ellison has carbon fiber racing yachts. I have a carbon fiber fiddle. I prefer mine... https://mastodon.social/@hyc/116195348554331559
mirabilos@mirabilos a distinct possibility, yes, but a win's a win.
@mirabilos still, LMDB was released in 2011 and here we are 15 years later and distros have only begun to stop shipping BerkeleyDB.
Postfix began supporting LMDB in version 2.11, released January 2014. They're only now telling users that it's time to migrate off BDB. https://www.postfix.org/announcements/postfix-3.11.0.html
Despite LMDB being orders of magnitude smaller/faster than BDB and completely crash-proof, it's been a long slow road.
Inertia is a helluva thing.
Kevin Karhan 