Quote from: Fuserleer on June 03, 2015, 04:27:55 PM

Quote from: bitcoinjesus2.0 on June 03, 2015, 05:25:11 AM

I know that 100,000 TPS sounds impossible, but so does an "automatic blockchain that pays for its own development with its profits"

100,000 TPS isn't impossible, its just impossible to do on a block chain in a true P2P decentralized manner without some form of centralization, or "magic trick". It seems to me BitShares may be doing both as per these 2 quotes in the link you provided.

"...assuming that all the witness nodes have internet connections capable..." so it is indeed confined/centralized to a set of nodes

and

"...with an average transaction size of about 100 bytes."

The latter concerns me, the absolute minimum core basic requirements of a secure verifiable transaction are a sender, a receiver, a value and a signature. Just the signature alone with a 256bit key is in the order of 70 bytes, sender pubkey is 32, receiver RIPEMD160 address is 20, and a value is 8 which is a total of 130 bytes with the bare minimum. If the key space is reduced to 160bit, then it will just fit in 100 bytes, but with a huge loss of security.

I'm assuming that to achieve this 100,000 TPS something similar to this happens:

Transactions are filtered through these "witness nodes" and I send 100 BTS to A. If within a certain time frame, A moves it to B, and B to C, etc etc that TX isn't recorded on the block chain until such a time that it lives at X for longer than a specified period of time. The 100 BTS movements between A -> B-> C....X are not recorded in full on the block chain (if at all), only the transaction A -> X

Taking that approach could indeed give you very high TX throughput, but if that is the method used (or something similar) it's a total hack in my opinion and may well lead to issues later.

Of course I'm speculating as I don't have the time to research this properly, so if you could provide some links/docs/something that details how this works rather than me hunting, I'd appreciate it. I stand by the fact that it is not possible to do, on a block chain, while recording all transactions to said chain, and allow any node to be a full node without special requirements.

Quote from: bitcoinjesus2.0 on June 03, 2015, 05:25:11 AM

Anyway, I'm getting off topic. 180,000 TPS might sound great to some, but Monero destroys BitShares in this arena with the best TPS ceiling of "infinity," so there's that option too of course. In other words, the bitcoin community has infinitely more options than they are currently looking at, and I am just trying to show them that the state of crypto circa 2015 is "not your dad's crypto"

If Monero really has stated "infinity" as their TPS limit, then someone there really needs a reality check! Regardless of what is possible on a block chain or not, the laws of thermodynamics will step in and dispatch a tough and thorough spanking waaaaay before "infinity" is even close

IMO the only way to achieve anything near a sustainable VISA level transaction throughput, stay in keeping with real decentralization (no special node sub-sets that are selected or voted), not perform any "tricks" which may compromise security, AND have all transactions on a public ledger is to scale horizontally, and NOT vertically!

Chain based ledgers can't scale vertically past a certain point, no matter how big your bag of tricks, nor your processing setup, horizontal is the only way and by that I mean a distributed and partitioned ledger of the ilk that we are doing over here. No one has even attempted to do this, because its assumed impossible or too difficult, and if it is so be it, at least it was attempted. However it's not impossible, we've ran it in many betas now and its is very close to being ready for use.

Quote from: bitcoinjesus2.0 on June 03, 2015, 05:25:11 AM

I like what you've done with e-munie. You should come work for the BitShares blockchain. Just submit a proposal, and the community would certainly vote you into a paid position (that's how BitShares members fund development).

https://bitsharestalk.org/index.php/board,61.0.html?PHPSESSID=2170a8f0b09b8fa2bdc7d35908ab4517

Heh thanks but no thanks, I've ploughed my life and everything I have into eMunie and I'm not jumping ship, ever

EDIT
----

So I did some more digging and came across this:

"...the idea being that if transactions have their signatures validated as they propagate across the network, a witness can have any number of computers surrounding him that validates all of these signatures, and then he gets a list of transactions and puts them in a block, and he doesnt have to check those signatures himself, because he has got all these other nodes surrounding him that are dividing up the task."

Can someone clarify this? Witness nodes, which build the blocks DO NOT check transaction signatures themselves, but rely on 3rd parties (which may be dishonest) to inform them that the signature for said transactions validate? How does a witness node know if a 3rd party is providing false information regarding a transaction, claiming that it contains a valid signature when it may not? If this happens, how then does the network resolve it, someone, somewhere must be doing a full validation of those 100,000 TPS to ensure that all transactions really are legitimate.

The issue is that sequencing and finalizing a block production is done single threaded, because validity is dependent on sequencing. That makes that step the bottleneck, since it can't be scaled to more threads. The signature checking doesn't have to be done by third parties, just separate threads. Those separate threads could be spread across a cluster, or possibly handled by GPUs, or anything else, they just have to be running somewhere trusted by the Witness in order to feed the final Witness thread valid transactions to sequence and include in the block.

It's always a trade-off between decentralization and therefore redundant block validation and transaction fees. Once you get to a globally useful scale, processing transactions has a real cost, and it has to be paid either by inflation of the currency, transaction fees, or subsidies from somewhere else. The BitShares approach is to let the shareholders decide how much redundancy and decentralization they're willing to pay for, and where to set the transaction fees based on that.

troglodactyl

Posts

Merits