Sorry but it was the first thing that came to mind 
3 Likes
We’re agreeing I think - should be a catch before there’s a problem in the first place, Victor is going to help us with a little more rigour - as clearly I for one, am an added liability 
3 Likes
haha - no it was me! Discord post confirming we will be making needed correction Monday. Thanks for leaning in with some tough but needed truths guys, it really is very appreciated - and even more valuable.
5 Likes
Just clarifying that I 
'd your reply for the additional rigor part, not the part about you being a liability. Yes it’s a mistake that shouldn’t have happened, and should have been automatically caught by tooling, so that’s a mistake, but the point of verification tooling like this isn’t because you’re not good enough. It’s because noone is good enough, including myself. That’s ultimately why Rust is so great, it conforms every line of code to rigor that a great engineer on their best day will do, but it does it consistently without fail.
4 Likes
Think your being a little harsh on yourself but can appreciate you taking one for the team 
6 Likes
We all know everyone is doing their best. The project is growing in complexity over time. It’s really important that our management takes breaks and doesn’t burn out.
5 Likes
Ha, that’s me all over, a happy accident 
7 Likes
Without emissions what incentive is there to keep a node running once the ‘pay once’ to upload has been received? And will emissions be enough to motivate the node operator to keep that data online permanently -esp where 100s of millions of nodes are competing for the tiny emissions?
The hit of the next upload payment…
6 Likes
‘Pay once’ is per piece of new data going onto the network, so if there’s a constant stream of new data being put on the network (new videos, photos, memes, backups, documents, social media posts, software versions, game mods, etc etc erc), there will be a constant stream of payments to node operators.
This stream of payments will be enough to incentivise enough nodes to stick around to match the demand from uploaders.
I don’t think so. Only demand for putting data on the network will be enough to sustainably inventivise nodes in the long term, and emissions are only planned for 12 years. After that, data payments alone will incentivise nodes.
5 Likes
This is why I think uploads should be paid even if they contain chunks that already exist on the network. I don’t think it’s designed to work that way though. The problem is, the more data that’s uploaded to the network, the less uploads will cost meaning less rewards to node operators. Taken to its logical extreme, every variation of a chunk that can exist on the network does exist, uploads are nearly free, and node operators earn close to nothing?
If I misunderstand I’d like to be corrected.
Most new data being created will be unique. I think this is the only thing you’re missing.
If I understand correctly, if 2 single-chunk videos were made identical other than 1 bit changed, they would be different chunks on the network & both be paid to upload. Only identical files, not similar files will be deduplicated.
Nobody will create a video / song / long document that is inadvertently a bit-perfect copy of another person’s file.
Trying to upload data that’s already on the network doesn’t need to incentivise new nodes to join, as no additional storage is needed for it, but if it uses other network resources (bandwidth, processing etc), then a proportionate fee to cover that would be sensible.
4 Likes
Initially, yes. Over time that reduces to zero, as every unique chunk that can exist on the network does exist. Just think of hash collisions.
1 Like
Correct, and all data uploaded is unique since dedup is caught by the client and any good uploader.
Maybe at the end of the universe. Currently the amount of new data is massively increasing. Every year there is so many times more (new) data created in the world. So one assumes some of that will get stored on the network.
Data 5 years ago in the world is dwarfed by the amount of unique data existing today. Go look up doubling of information. And be amazed, I can say that since you said
Now as to hash collisions, well there are more unique hashs then there are electrons in the known universe. So doubt collisions will be occurring very often at all.
5 Likes
The unique space is fixed. It’s the unique byte variations in the capacity of a single chunk. Data increasing only underscores my point.
Hash collisions was only to analogize my point that “real big numbers bro” is hand wavy. There will be overlaps, and over time that’ll increase.
My point is that over time, as the network has more and more data uploaded to it, less and less of that data will be a chunk that doesn’t already exist, and require payment.
Yep, and its way more than the atomic particles in the known universe.
If you generated a truly random number at the rate of 1,000,000,000,000,000,000 per second then there is 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 seconds before you create them all and just to get a random match its on the order of 1,000,000,000,000,000,000,000,000,000,000 seconds at least
Not happening any time soon, like a million years since new unique data is doubling in such a short time and that time is still reducing.
2 Likes
Can you give an example of the kind of data that will be uploaded that is not unique & will increase over time relative to unique data?
E.g. if people creating and uploading videos were a big use of the network, there will never be a time when 2 people make the exact same video if they’re recording their own voice & a unique event.
Also, there will never be a time when people upload enough video chunks to cause frequent hash collisions. ‘Really big number’ isn’t a hand wavey answer, it’s real statistics.
If somehow there was a way this became an issue in the distant future, I’m sure the network could adjust the addressing to cope.
2 Likes
shrug heard the same thing about hash collisions. There’s a difference between calculations assuming perfectly spread data (no overlaps until all slots are filled) and what happens practically. (could have an overlap day 1)
Well it’s not about files, where different sizes factor in. It’s about chunks which are uniformly sized and not that large.
1 Like
If you noticed I gave the expected time for a collision. Which is smaller than the time to generate all.
Tell me, search for it, how many BTC address generations that have caused a collision? And that is a smaller hash size
The hashes in Autonomi are 256 bits long (approx 10^77 possible values). More atoms in the known universe.
Would you agree that over time, the odds of an overlap increases?
If overlaps aren’t paid for, then there’s an economic factor at play that I haven’t seen addressed.
Take a successful Autonomi future in mind. There’re no emissions after 12 years, upload fees are all that goes into compensating node, and those dwindle as overlaps increase.
I think it would be better for the network to charge for the size of the upload whether there are overlaps or not. Day 1, uploading a 1GB file (assuming it worked 
) requires paying to upload 1GB, whatever that cost is. Day 9 jillion, uploading a 1GB file requires paying to upload 1GB, whatever that cost is.
1 Like