Excellent point, it is easy to forget the minor miracle of FOSS. Plus, without the ad revenues there would be no mega-corps vacuuming up all the new grads, so I would anticipate a significantly greater rate of innovation and FOSS contribution broadly.
There's even an interest or social-reward factor to participating in these kinds of things. Working on an open-protocol messaging client for free is a lot less rewarding when the userbase of the entire protocol is 1% or less of all online messaging, because most of that market's captured by closed platforms that forbid and/or discourage other clients, than when it works with 20+% of clients and even your non-geek friends are using the protocol, if not your particular client.
I truly think we couldn't launch something like the email protocol these days and have it gain traction, and I don't mean because of its flaws. I judge that a pretty crappy state of affairs, and I think the #1 cause is that it's so lucrative to keep your users in a position where you can track & spy on them very well, while avoiding leaking anything they're doing so that competitors can see it—IOW incentives are set up to greatly reward successful closed platforms while discouraging interoperability, so we get even more of that than we otherwise might.
> I truly think we couldn't launch something like the email protocol these days and have it gain traction...
Sadly it's worse than you expect here. Enter ElasticSearch.
The company behind the innovation you propose will piggy-back on open source projects (Lucene), add a novelty to it (clustering) and choose a permissive open-source license to encourage contributions. Once hitting a significant market penetration threshold, the project then will move to a mixed-source, enterprise license model with intentionally-crippled community versions (think Neo4j, JFrog, etc).
ElasticSearch isn't even alone here, it's just the most obvious example. I've actually been insisting for a long time that we need an Apache-licensed standard solution for clustering generic applications...something useful enough that anyone can connect part A & part B to get "clustered Lucene" instead of "ElasticSearch". Something reasonably deployable (read: is monitorable, has RBAC) without massive licensing costs (read: Neo4j). Not an easy problem, for sure.
