Swirlds - Hashgraphs as a promising alternative to Blockchains?

[alkan]

I just came accross Swirlds and its concept of Hasgraphs is fascinating as it could solve practically all the issues of existing cryptocurrencies like Bitcoin and PoS coins.

http://www.swirlds.com/downloads/Overview-of-Swirlds-Hashgraph.pdf
http://www.swirlds.com/downloads/Swirlds-and-Sybil-Attacks.pdf
http://www.swirlds.com/downloads/SWIRLDS-TR-2016-01.pdf (Complete White Paper)
http://sammantics.com/blog/2016/7/27/hashgraph-consensus (Q&A)

It seems that a cryptocurrency based on the Swirlds framework could not only cope with high transaction rates (like VISA) due to its nearly-optimal efficiency, but Swirlds also allows the user to discard the old history of the hashgraph. According to the makers, that would shrink the amount of storage from Bitcoin’s current 60 GB to a fraction of a single gigabyte.

As an underlying framework, Swirlds could work together with PoW, PoS, PoB and other systems. If I understand it correctly, it would be possible to build a a PoS coin with two fundamental advantages over existing altcoins:

1) The prococol allows objective consensus, i.e. a new node could independently arrive to the same current state as the rest of the network based solely on protocol rules and messages propagated across the system. In contrast to regular PoS coins, there's no longest chain-rule and no blockchain that an attacker could forge at will. Instead, the network is based on locally stored hashgraphs containing all the "gossip about the gossip" about the actual transactions.

2) Nothing at Stake attacks aren't possible since there's no blockchain that could be double-mined on. In fact, forks are part of the system that is based on DAGs rather than chains.

[1714785552]

1714785552

[1714785552]

1714785552

[1714785552]

1714785552

[spartak_t]

This sounds very ambitious, but there are only few demo's shown and the whitepaper (from what I've read so far) does not provide information on how they will deliver the technology.

[alkan]

This sounds very ambitious, but there are only few demo's shown and the whitepaper (from what I've read so far) does not provide information on how they will deliver the technology.

Well, the whitepaper in the link above is very detailed and describes the algorithm in depth, providing mathematical proofs for the claims.

[alkan]

If you're interested in a short introduction to Hashgraphs, you should check out this presentation:
https://www.youtube.com/watch?v=rhHfjOqlUv4

The guy seems to really know what he's talking about. Just look at his list of publications: http://leemon.com/papers/pubsTopic.html.

However, there's one thing that appears like an issue to me. An that is the importance of the number n of all the nodes, which is a fundamental variable in the calculations done by the algorithm.

Do all nodes need to be online for system to function?   Number of current nodes?

Over 2/3 of the nodes need to be online for consensus. If fewer are online, the transactions are still communicated to everyone online very quickly, and everyone will immediately know for certain that those transactions are guaranteed to be part of the immutable ledger. They just won't know the consensus order until more than 2/3 come online.

Especially in a non-permissioned system, the number of nodes will vary considerably over time. How will n be determined in such a system so that it not only allows for communicating existing transactions but also for consensus on new ones?

It's not easy to determine (or even estimate) the number of nodes in a p2p nework that are online at a given time. And even if you could guess that number, what would be the effect of a highly varying n on the hashgraph itself? Or would you first need to reach consensus about the size of the community?

This is what Leemon seems to imply and according to him, this would work out nicely (https://www.youtube.com/watch?v=NI6sQHfIPbk, starting from 29.15). But isn't that rather circular reasoning? Or maybe you could just include the fact that a node isn't reachable anymore in your gossip. If more than 2/3 of the nodes gossips that node x is down, n would be reduced by 1 by all nodes as per the consensus achieved by them.

[Spoetnik]

Hmm interesting.
I have to admit i am a bit lazy / busy to sift through White Papers right now though.

Are they (who ever they are) providing the concept only or do they plan to implement and release it functional ?

[alkan]

Are they (who ever they are) providing the concept only or do they plan to implement and release it functional ?

They are providing a SDK (and some demo projects like a simple stock market and a very simple game) which allows you to build all sort of applications on top of Hashgraphs.

A video tutorial on the SDK can be found here: https://www.youtube.com/watch?v=NI6sQHfIPbk

[spartak_t]

There is no information on how they are planning to distribute it though. Only e-mail for contact is available.

[alkan]

There is no information on how they are planning to distribute it though. Only e-mail for contact is available.

Once you have filled in the form on http://www.swirlds.com/download, you can download the SDK instantly.

But what really matters (to me) is the core idea rather than the implementation of it. If somebody is interested in building a system based on Hashgraphs, all he needs is the whitepaper.

[bwang22]

Really interested in this. Not sure why they haven't create a cryptocurrency like IOTA or Bytaball has based on DAG. Maybe swirlds isn't interested or is Hashgraph just not suited?

[ripbit]

there's one thing that appears like an issue to me. An that is the importance of the number n of all the nodes, which is a fundamental variable in the calculations done by the algorithm.

Do all nodes need to be online for system to function?   Number of current nodes?

Over 2/3 of the nodes need to be online for consensus. If fewer are online, the transactions are still communicated to everyone online very quickly, and everyone will immediately know for certain that those transactions are guaranteed to be part of the immutable ledger. They just won't know the consensus order until more than 2/3 come online.

Especially in a non-permissioned system, the number of nodes will vary considerably over time. How will n be determined in such a system so that it not only allows for communicating existing transactions but also for consensus on new ones?

It's not easy to determine (or even estimate) the number of nodes in a p2p nework that are online at a given time. And even if you could guess that number, what would be the effect of a highly varying n on the hashgraph itself? Or would you first need to reach consensus about the size of the community?

This is what Leemon seems to imply and according to him, this would work out nicely (https://www.youtube.com/watch?v=NI6sQHfIPbk, starting from 29.15). But isn't that rather circular reasoning? Or maybe you could just include the fact that a node isn't reachable anymore in your gossip. If more than 2/3 of the nodes gossips that node x is down, n would be reduced by 1 by all nodes as per the consensus achieved by them.

Alkan, I had the chance to ask Mance Harmon, Swirlds CEO, your question on the Hashgraph Telegram Group:  https://t.me/joinchat/DpeJM0R1FpqUD5N2YvNcAQ

This was his response:

"What's important to remember is that the hashgraph is unlike other gossip protocols.  The information that is being gossiped is both transactions (understood by the application running on top of the platform) and information that conveys who has talked to whom and when.  So, it's incorrect to say that "It's not easy to determine (or even estimate) the number of nodes in a p2p network that are online at a given time."   Actually, it's incredibly easy if you have a hashgraph that represents Gossip About Gossip.  It's easy to look in the hashgraph and see the last time a given node gossiped.  It's directly observable in the hashgraph.  The hashgraph actually gives you a fantastic view of network weather and who is online and when.

In a permissioned system we only let people join if they are going to be active.  Because the hashgraph comes to a moment in time where all nodes agree on exactly when a transaction takes effect, they can all agree on when the number n changes because somebody joined or left.  So, at every moment every node is in agreement on the value of n.

Today we are describing permissioned systems.   We will describe our ideas for public systems in the future."

I hope that helps.  If you join the telegram group you can chat direct with some of the lead developers Hashgraph Telegram Group:  https://t.me/joinchat/DpeJM0R1FpqUD5N2YvNcAQ

[hashgraph]

Hi friends ! Just wanted to stop by and make sure you got the correct Telegram link.

http://t.me/hashgraph

In addition to www.swirlds.com we also have www.hashgraph.com

Dr. Leemon Baird (inventor of hashgraph) just gave a talk at CESC2017 at Berkeley on “why math matters” and the importance of Asynchronous Byzantine Fault Tolerance, which I think you will enjoy. https://youtu.be/c6_AXwAcUnU

Talk soon.

[multinova]

This tech really interests me.. not much traction for now.. I hope to see it adopted soon.

[tech72]

Do they sell any coin or something like currency or dapps ?

[multinova]

Do they sell any coin or something like currency or dapps ?

Not yet as far as i know. Seems to be all in pre release phase.

[flag39]

Hi friends ! Just wanted to stop by and make sure you got the correct Telegram link.

http://t.me/hashgraph

In addition to www.swirlds.com we also have www.hashgraph.com

Dr. Leemon Baird (inventor of hashgraph) just gave a talk at CESC2017 at Berkeley on “why math matters” and the importance of Asynchronous Byzantine Fault Tolerance, which I think you will enjoy. https://youtu.be/c6_AXwAcUnU

Talk soon.

Could you make a comparison Hashgraph vs Tangle? 

[sanjoant]

Tangle , Hashgraphs seems necessary improvements to over come  existing inefficiency of pow/pos blockchain architecture. Waiting to see  this evolve in to a real life use case with token.  

[hashgraph]

Both hashgraph & tangle use a DAG (Directed Acyclic Graph), but other than that there is little similarity.

Tangle has a PoW component, hashgraph needn't.

hashgraph has virtual voting, tangle doesn't etc etc

We will be creating a more thorough comparison/analysis and revising the hashgraph FAQ accordingly.

[plauthor]

It seems to address blockchain's shortcomings. We have been therefore considering implementation of the Parsable Law project in hashgraph rather than in blockchain. The project's draft white paper is now available for comments:

https://lookaside.fbsbx.com/file/masek_zdenek_parsable_law_draft_november_2017.pdf?token=AWxyK6XgIZwDB67oDi-NLnID2RC6sNOkSrIB7DT0GvW_tn4pIrZcTcl-WBMbnJzbyzsArsl1XSfuEDwjPmSLRqt2r9YYjCnKI3ZAzoTF-MARLi5NId-Ngt--GUqiJV9eR2mPkZqTlNNeZzGVKFKP-Sq8Vjs9jn9IqtPL3jGpaeWGI2da58iM2jszxSzzz68f67z58ePoUzskiCJShW_v2bP4fnZ6UysaQVK8dSF8ZhgdGQ

[13Darko]

How does Hashgraph solve an issue when an attacker creates 2 conflicting transactions which after a bunch of steps through different nodes will be met in the same node? Can this bring us to a deadlock?

[multinova]

Is this actually the official ANN thread of Hashgraphs ?
Or is there another one available for more questions ?