The Simple Signal Sharing Protocol is an interactive application-level protocol making use of Ethereum smart contracts and the WebSocket standard for streaming signal data over a period of time between two participants; a signal streamer and a signal consumer.
A trustless payment mechanism is the core of the protocol; allowing the signal streamer to gradually monetize their their signal and the signal consumer to receive access to novel data streams.
Data sharing over the Internet has been made out to be technically challenging by a few emergent companies e.g. Ocean Protocol, streamr and others. Key challenges include the pricing of a data set and the trustless and atomic swap of payment and data. Similarly, so called "data escapes" have been portrait as "deal breakers" in the space too.
Having taken a look at some approaches to data sharing; we think there's still amazing potential to improve the space with actual technical innovation. In the following document; we hence like to introduce a simple Ethereum-based protocol that allows to parties to share data verifiably and gradually - hence reducing the risk of either party betraying the others.
All anonymous data sharing protocols currently require two parties to trust each other. Indeed, totally anonymous sharing of data that involves crypto currencies is nearly impossible without accepting the risk of one party betraying the other.
For the scope of this work; we'd like to highlight two paths where data sharing can fail in traditional protocols.
(1) Assuming that the data consumer pays the data publisher upfront; then the risk of the data publisher just sending a random set of zeros and ones is high and irretrivable. If e.g. on Ocean Protocol, I transferred some coins to a data publisher and then proceeded to download the data set; I'd have no option for recourse considering that anonymous users can enter data sets. Now, I'd have lost money and didn't receive any data.
(2) Similarly, in the case that the data consumer pays the data publisher after receiving the data; then the risk of the data consumer exit scaming the data publisher is high and irretrivable. If e.g on Ocean Protocol, I downloaded a data set from a publisher but then simply forgo payment afterwards - then the data publisher would have no option for recourse considering that I could have do all previous actions perfectly anonymous using e.g. the TOR network. Now I'd have the data set and it wouldn't have cost me any money.
Being aware of these two scenario makes one realize that, in fact, the problem isn't an easy one to solve atomically either as we'd argue it's rather difficult to understand what equates a valuable data set.
Humans are fundamentally not good at understanding large corpuses of data and so in cases where e.g. users download data sets; they may end up not being aware of its quality until they've done some tests.
Hence, a key lesson that lies within the scope of this work, is that "shipping data sets in objectified containers" isn't the right conceptual metaphor for tackling the over all problem.
Indeed, data is often portrait as lake, package, container, zip file, csv etc. But the reality of anyone working with data is that any point that makes a set of data can be viewed as its own self-contained set of data too. Hence, in future paragraphs of this work, we tend to favor the term "stream" or "signal" over data as they correctly identify data as time-continuos resource rather than one that is akin to an object.
A smart contract hosted on an EVM-enabled platform like Ethereum allows a data publisher to offer a signal stream by publishing some basic meta data, e.g.:
{
stream: {
name: "ETH/USD"
}
}
A digital signal is used to represent data as a sequence of discrete values. In theoretical terms, a signal is modeled within a vector space , where represents a function's domain. Commonly-used domains to model digital signals include to model time series data, or to model signals dependent on location [1]. Within the scope of this work, we assume that a digital signal's domain is descrete, hence representing a finite number of isolated points.
Sampling is the process of reducing a continous-time source signal into a discrete-time replication signal by repeatedly applying inputs to the source.
Aliasing is an effect appearing in signal sampling. It can lead to a replication signal becoming indistinguishable from its source [3].
Shannon's theorem states that if a function contains no frequencies higher than hertz then it can be determined by polling a series of points spaced seconds apart. It can, hence, be assumed that taking more than samples of a function with a frequency no higher than hertz yields a perfect reconstruction through sampling [2].
Asynchronous systems are able to handle the program flow as well as the occurrence of independent events without the need for a a global coordination mechanism like a clock or a centralized consensus. They do not rely strictly on arrival times of messages or their ordering.
A system is stateless when it treats a any occurring message independently and without the need for keeping track about messages in memory.
The Simple Signal Replication Protocol is an application-level protocol with the lightness and speed necessary for distributed and collaborative information systems. It allows the replication and discovery of digital signals over the Internet.
The following shows an example of a source initializing the signal replication
by replacing (PUT) the current entity (BTC-USD) with an incomplete csv
content. Using PATCH, the source then partially modifies the entity through
additional data.
PUT /signals/BTC-USD HTTP/1.1
Host: destination.com
Content-Type: text/csv
timestamp,value,currency
2021-02-19T22:37:26.102Z,55,529.65,USD
2021-02-19T22:39:09.973Z,55,550.11,USD
PATCH /signals/BTC-USD HTTP/1.1
Host: destination.com
Content-Type: text/csv
2021-02-19T22:37:26.102Z,55,529.65,USD
2021-02-19T22:39:09.973Z,55,550.11,USD