Ignore this
I’m not going to sit here and say this is the best way to design things, but for one reason or another, it’s the way I’ve always done it. As a matter of fact, my guitars involve a lot of improvisation too — so does my cooking.
This means, of course, that when it comes to learning, I’m firmly in the camp of those who eventually do. You see, the path I always seem to take is littered with landmines, and although I’m well aware of this, I remain remarkably proficient at stepping on them anyway.
I only preface this little reflection because today I think I had one of those moments. As some of you know, I’ve been helping out the folks at @threespeak for months now, and while things have been working fairly well, they could definitely be better.
Speed is an issue — and as far as I can tell, it will continue to be an issue with the current infrastructure. The mètre des archives has to be YouTube, and I say that without joy. Whether we like it or not, whatever we build has to move us closer to the giant.
With that in mind, I keep coming back to the fact that here on Hive, people don’t really watch old videos very much. I’m sure some do, but most videos capture their audience in the first day or two, then slowly fizzle into the cold-storage corners of the InterPlanetary File System — IPFS for short.
So, in the same way we came up with staging servers for uploads, and fallback systems for when the IPFS supernode is feeling cheeky, maybe there’s one more move we can make. A move that could significantly improve the overall experience.
You see, the IPFS supernode is a monster — but it’s not the fastest of monsters. Yes, it can hold insane amounts of video (as it currently does), but the redundancy required for safety also makes it less than optimal at times.
If encoders were to first upload videos to a staging video server — one equipped with blazing-fast NVMe drives and a Mach-5 internet connection — even if only for seven days, that would be a game changer. That’s the TL;DR.
This staging server wouldn’t need massive storage capacity. Its blocks would constantly rotate, migrating to the supernode for long-term storage as newer content comes in.
At any rate, this might be the last missing piece of the puzzle — the one I hadn’t considered until today. I’ve already shared the idea with Eddie, and I think we both agree it has real potential.
That said, decisions like these — and the actions that follow — aren’t mine to make, nor are they particularly timely. With Hive prices where they are, adding infrastructure costs feels like trying to save a newborn while swimming in subzero waters.
— MenO
▶️ 3Speak