boltstream VS Flythrough.Space

Several years ago I cobbled together all the pieces to make a live video streaming website [1] like Twitch.tv, et al. It's impossible to actually run at any kind of scale without being an ISP or Google-scale peer. Not to mention the enormous amount of DMCAs when people are trying to stream live sports, Pay-per-view, etc.

It was fun to play with it for a couple months. But there's no business here. Twitch.tv is also finding that out.

[1] https://github.com/benwilber/boltstream

I built Boltstream [0] over the course of a couple years and got burned out. I wouldn't say it was actually a failed project because in the back of my mind I never really had any intention of trying to turn it into a Twitch/YouTube/Facebook Live competitor anyway. And it's proven to be somewhat popular on Github. The project was fun to build and learn about but ultimately it was impossible to actually run it as a consumer-facing website. The problems were never the actual software/video streaming tech, but rather the extreme bandwidth costs of live video streaming, and the content itself. I launched it several times under different site names/products and it always just turned into a cesspool of pirate streaming live sports and other copyrighted content. The DMCA notices were regular, but were never really that onerous to deal with. Just turn the stream off, ban the account, and then reply to the email. But it just got too annoying. I built a little mobile web app so I could just do it from my phone while I was out at dinner. After awhile I just decided that it was time to give it to someone else to play with. So I did a big code dump on Github and haven't touched it since.

I've actually started working on it a little bit more recently since it seems that "self hosted video streaming" is still pretty in-demand. I'm probably not going to be spending too much time on the actual features and functionality since that's pretty much done at this point. Mostly just packaging it up so it's easier for people to run it themselves and hack on it.

[0] https://github.com/benwilber/boltstream

Here's what it looks like now (planet files):

My project was a 3d in-browser escape velocity clone (think Endless Sky.) It ended up with an impressive feature set, mediocre graphics, not nearly enough world, no real game loop, and no players: http://flythrough.space

My mistake was building it totally in secret for most of its life. I worked on it for years focusing on adding features that maybe nobody wanted without validating the product/market fit or generating any buzz. By the time I was "ready" someone had gained way more steam on their own EV remake, even getting the author of Override to endorse it, so the fact that I had a working prototype didn't really matter, the community that I thought would be receptive wasn't interested.

I learned that you need to build something worth playing first, including some kind of game loop, before you try to get people to play it. Turns out noodling around with a hacky prototype isn't something people find exciting even if the whole community is centered around noodling around with mods and hacking. People won't mod a game they don't love in the first place.

Since this project, I've committed to building smaller prototypes and validating concepts before I go all-in on building something polished and content rich, and thought I haven't shipped anything this big since, I have met my personal goals.

Full retrospective: http://blog.eamonnmr.com/2020/08/flythrough-space-retrospect...

The conclusion: http://blog.eamonnmr.com/2020/04/dont-remake-an-old-game/

Here's a real example from a project I worked on in my homebrew ECS framework and later implemented in Godot: a guided missile.

In the ECS project the notion of 'thing that can run into other things and do damage' so totally separate from the notion of 'thing that is driven around by AI.' So adding guidance to an existing projectile wasn't too much of a pain. In the Godot project I'm dealing with networking as well, so the division is between fire-and-forget projectiles (derived from Bullet) which know about hitting things and doing damage, and AI driven Ships with AI or player driven movement which needs to be synced over the Network. In that case I had to copypasta the AI driven movement code into a 'guided' class under Bullet. That said, for almost every other task, Godot's composition first model has been way easier to work with, especially because it lets you test elements in isolation. Here are the two projects if you'd like to compare the code:

Homebrew ECS framework on top of the babylonjs engine: https://github.com/EamonnMR/Flythrough.Space

Godot with networking: