1. 57

  2. 22

    It being new and proprietary tech is the particular killer. It would still be disruptive if it were open-source, but in that case you at least have the out, if enough $$$ are riding on it, to hire a team to keep maintaining it for you (maybe even some of the old team).

    That can also be done with proprietary products, but seems to only be common among big traditional engineering firms. In that field, big companies often negotiate a license that includes private source code access and the right to maintain an in-house version indefinitely, so their own stuff can survive death of the vendor.

    1. 18

      I hope some of the folks at datomic, hyperdex-warp, neo4j, take some time to explain their position on this. Not trying to rabble rouse, but I’d just like to know how they do their licensing to protect you from this apparent nightmare scenario. datomic mentions ‘perpetual license’ but it’s confusing how that works after the support is up. neo4j is subscription. Hyperdex is half open, so maybe they can capitalize on the loss of foundationDB to make some sort of open move on their ACID (warp) functionality.

      Interesting business opportunity here, but as per the title don’t bet your new business on unproven, non-replaceable tech if possible.

      1. 5

        Neo4j is GPL3/AGPL3, but the free software version doesn’t support clustering (“high availability”) or hot backups. This is probably problematic for a lot of companies but at least if your use case doesn’t require extreme uptime you’ll definitely have time for an orderly migration in the event Neo Technology gets acquired or otherwise stops offering its product for sale.

        1. 2

          Thanks! I also noticed that a founder of aerospike (which went open source) was commenting on the HN thread on the foundationdb situation. I just think it’s an interesting place for other players to move in.

        2. 3

          I asked Cognitect (the makers of Datomic) about this and their support chap said that they offer code escrow for Enterprise licenses if desired. (It doesn’t say so explicitly on their support page, but I suppose it is implicit in the third row of this table on their support page.)

        3. 7

          IMO, a reasonable rule of thumb is that any technology that is in the customer flow should either be open source (preferably with internal competence) or built internally completely. Most other options are a liability.

          Things like Payroll are probably fine being bought as downtime there might make your employees unhappy but (probably) won’t bankrupt your company.

          1. 5

            If a company I worked for failed to pay me, even if it was a payroll downtime issue - I would be looking for a new job asap.

            1. 4

              I make some exceptions for tools that are easily replaceable. I’m currently betting on a new datastore, but it’s fed via a standard protocol (graphite wire format). If I decide to ditch that portion it won’t be very hard because none of the inputs or outputs will have to be changed (although I will have a few hundred gb of data to move).

              1. 2

                Open source software or open protocols (with more than one implementation). It’s a good rule of thumb for everything, not just databases.

              2. [Comment removed by author]

                1. 7

                  I’ve seen a company bankrupted when Oracle decided they wanted more money, thanks. It’s not at all a safe bet. I would not want to build a business unless I was confident that we could, in extremis, switch to a competitive alternative.

                  1. 2


                    A rule of thumb is a principle with broad application that is not intended to be strictly accurate or reliable for every situation.

                2. 10

                  Yes and yes - although many people learned this lesson long ago with MongoDB, which had nothing to do with being bought out.

                  1. 7



                    If data have structure store Postgres. If data no have structure store Hadoop. If data no have value store Mongo.

                    24 May 2013

                    1. 14