Re: Blockchains and the Web of Things?

Hello,

On Tue, Aug 30, 2016 at 8:06 PM, Drasko DRASKOVIC
<drasko.draskovic@gmail.com> wrote:
> This is very kind of you. I will give you a call, an then re-paste
> here the meeting minutes.

I had extremely interesting Skype meeting with Tibor who was very kind
(and patient :)) to explain me lot of things about his Blockchain
visions and how they propose the solution in Streembit project
(http://streembit.github.io/).

As promised, here are (quick and dirty) meeting minutes (more in the
form of Q&A):

[Q] Blockchain is distributed ledger. How do you stock the list of
transactions in IoT network, having in mind that is it comprised of
highly constrained low-power devices
[A] Full Bitcoin node can reach today 90GB. Even on PC this becomes
impractical. Partial solutions are trusted nodes. **Only gateways**
store data in DHT - gateways are powerful nodes - RPi can have this
role for the demo. Devices behind the gateways do not run Kademlia
DHT.

[Q] What are the benefits of decentralization if you have "partialy
decentrlaized" network, i.e. some kind of private subnets?
[A] There are many - some are described here:
http://streembit.github.io/2016-03-02-Internet-of-Things/ and here:
http://streembit.github.io/2016-07-15-Streembit-IoT/ - like device
discovery, distributed storage for firmware update, etc. Price is
lowered dramatically (no centralized expensive infrastructure).
Security is equally important, as centralized server is a target for
hackers, industrial espionage, etc - we are distributing the risk.

[Q] Do we need Blockchain at all? DHT is enough for device discovery.
[A] We need some kind of distributed ledger as we need certifications
and device audits. Simple merkle tree ahould do the job.

[Q] About the storage of messages needed for data analytics, machine
learning, big data... Is it distributed now?
[A] Yes - this also - beside identity management - is the role of DHT
- it keeps distributed storage of all historical messages (can be set
per node, etc).

[Q] What about micro-transactions - do they make sense? Like machine A
buying service from machine B?
[A] Experiences tell that this is not really feasible at this moment.
Should not be neglected though. But challenges are huge - Bitcoin has
delay of 10-15 min (6 blocks must be processed). Ethereum about 40-50
sec. But for IoT we need **instant** transactions (and for sure all
centralized solution like PayPal, Visa, Google, Apple - all of them
can allow this). This is practically impossible to solve with current
Blockchain. Other big problem is 51% or Byzantine Generals attack
(this is where the delay comes from in the first place) - how will we
protect the network.

Tibor,
I have surealy forgotten or misinterpreted something, so please feel
free to correct.

Thank you again very much for your time and clarifications.

Best regards,
Drasko

Received on Wednesday, 31 August 2016 20:13:43 UTC