Bitcoin Forum
November 24, 2017, 12:06:03 PM *
News: Latest stable version of Bitcoin Core: 0.15.1  [Torrent].
 
   Home   Help Search Donate Login Register  
Pages: « 1 2 3 4 5 6 7 8 9 10 11 12 [13] 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 ... 143 »
  Print  
Author Topic: Tau-Chain and Agoras Official Thread: Generalized P2P Network  (Read 259878 times)
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 26, 2015, 03:16:01 AM
 #241

i like the idea, something new at least.

I like the original distributed computer idea. Wonder where that went?

It never went but is only one among many other features in Agoras, most of them (which I try to put emphasis on) are possible only thanks to tau's unique features.

Tau-Chain & Agoras
1511525163
Hero Member
*
Offline Offline

Posts: 1511525163

View Profile Personal Message (Offline)

Ignore
1511525163
Reply with quote  #2

1511525163
Report to moderator
1511525163
Hero Member
*
Offline Offline

Posts: 1511525163

View Profile Personal Message (Offline)

Ignore
1511525163
Reply with quote  #2

1511525163
Report to moderator
1511525163
Hero Member
*
Offline Offline

Posts: 1511525163

View Profile Personal Message (Offline)

Ignore
1511525163
Reply with quote  #2

1511525163
Report to moderator
Join ICO Now Coinlancer is Disrupting the Freelance marketplace!
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1511525163
Hero Member
*
Offline Offline

Posts: 1511525163

View Profile Personal Message (Offline)

Ignore
1511525163
Reply with quote  #2

1511525163
Report to moderator
1511525163
Hero Member
*
Offline Offline

Posts: 1511525163

View Profile Personal Message (Offline)

Ignore
1511525163
Reply with quote  #2

1511525163
Report to moderator
1511525163
Hero Member
*
Offline Offline

Posts: 1511525163

View Profile Personal Message (Offline)

Ignore
1511525163
Reply with quote  #2

1511525163
Report to moderator
fartbags
Legendary
*
Offline Offline

Activity: 1176


View Profile
August 26, 2015, 06:09:50 PM
 #242

The marketcap is 7 million USD. The project looks cool but I'm not sure if this should be valued at more than 1 million dollars right now today.
Can you make some sort of comparison to other projects and valuations?
I have no idea what this even does. I've never seen a demo video of what you can do with TAU.

No video yet, but did you catch www.idni.org/blog/code-and-money ?



Don't you have the majority of TAU? If you consider TAU money, then you have it all. Now do some code! LOL



ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 26, 2015, 06:23:56 PM
 #243

The marketcap is 7 million USD. The project looks cool but I'm not sure if this should be valued at more than 1 million dollars right now today.
Can you make some sort of comparison to other projects and valuations?
I have no idea what this even does. I've never seen a demo video of what you can do with TAU.

No video yet, but did you catch www.idni.org/blog/code-and-money ?



Don't you have the majority of TAU? If you consider TAU money, then you have it all. Now do some code! LOL




have a look at the code github.com/naturalog/tauchain

Tau-Chain & Agoras
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 27, 2015, 06:24:06 AM
 #244

A special offer to wholesales in september was announced: http://www.idni.org/blog/sep-special

Tau-Chain & Agoras
klosure
Jr. Member
*
Offline Offline

Activity: 49


View Profile
August 27, 2015, 06:44:19 PM
 #245

I have a few questions:
- Do you have a specification (ontology, BNF grammar, spec document...) of the Tau language as it is now and/or as it is supposed to become once all the networking and crypto features are built in?
- Where can I find a Tau code example that emulates simple procedural behavior like a loop for instance?
- Will the Tau node to be a full fledged RDF triple store and SPARQL end-point?
- How do you plan to make the transition from knowledge (decision of spawning a new socket for instance) to action (making a system call to create a new socket): will each i/o module poll the knowledge base periodically for relevant action? Will you place control points at key steps of the inference process to intercept propositions with a known action semantic and take action?
- Will input from i/o module trigger asynchronous reasonning or will it just add new propositions in the knowledge base for some synchronous process to reason with at next iteration?
- Is the interface between the knowledge base and the system level functionalities already developed or specified? Where can I find further details?
- How can Tau guarantee that it can determine the upper bound of when a program can halt in less time than it would take to run the program? What if the program is already the shortest possible program to solve the problem it represents? At most I could imagine how code analysis could help determining a big-o estimate of the program complexity but then that won't help much to make sure that the program can run within the time boundaries of a given blockhain block as was suggested in the cointelegraph interview.
- Where can I find more (code / documents) about the JIT compilation of the Tau language?
- Are there archives of IRC discussions / forum discussions / email discussions etc that can help understanding decisions made around the design?
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 27, 2015, 09:02:36 PM
 #246

I have a few questions:
- Do you have a specification (ontology, BNF grammar, spec document...) of the Tau language as it is now and/or as it is supposed to become once all the networking and crypto features are built in?

The syntax of the language itself is not new. The RDF family is basically all syntax-sugaring to NQuads http://www.w3.org/TR/n-quads/ so all RDF languages can be converted to NQuads, while plenty of converters out there. Current tau's code can read NQuads, JSON-LD, and (yet partial) Notation3.

Supporting network features etc. is done via builtins, which is also a common mechanism among RDF reasoners. See for example http://www.w3.org/2000/10/swap/doc/CwmBuiltins

- Where can I find a Tau code example that emulates simple procedural behavior like a loop for instance?

Let me use the cwm builtins I just linked to demonstrate printing numbers from 1 to 10:

1 log:outputString 1.
{ ?x log:outputString ?x. (1 ?x) math:sum ?y. ?y math:lessThan 11 } => { ?y log:outputString ?y }.

see http://pastebin.com/H2gqLUzC

- Will the Tau node to be a full fledged RDF triple store and SPARQL end-point?

Except I won't promise now we'll use exact SPARQL syntax, then sure. Tau can be equivalently described as a decentralized triplestore with dependent types.

- How do you plan to make the transition from knowledge (decision of spawning a new socket for instance) to action (making a system call to create a new socket): will each i/o module poll the knowledge base periodically for relevant action? Will you place control points at key steps of the inference process to intercept propositions with a known action semantic and take action?
- Will input from i/o module trigger asynchronous reasonning or will it just add new propositions in the knowledge base for some synchronous process to reason with at next iteration?

At the syntax level, IO is just another builtin predicate, like { ?z a tau:tcpEndpoint } => { ?y tau:tcpConnect ?z }. At the typesystem level, we of course have Effect types/Monads https://en.wikipedia.org/wiki/Monad_(functional_programming)
The reasoning flow does not change at all due to IO (in general), but types originated from IO are tagged as effect types and of course aren't treated as "pure truth". A good reference is how Idris use side effects http://eb.host.cs.st-andrews.ac.uk/drafts/eff-tutorial.pdf

- Is the interface between the knowledge base and the system level functionalities already developed or specified? Where can I find further details?

Apologies but I didn't really understand the question.

- How can Tau guarantee that it can determine the upper bound of when a program can halt in less time than it would take to run the program? What if the program is already the shortest possible program to solve the problem it represents? At most I could imagine how code analysis could help determining a big-o estimate of the program complexity but then that won't help much to make sure that the program can run within the time boundaries of a given blockhain block as was suggested in the cointelegraph interview.

This goes right back to decidability, and to the point that one shouldn't expect the autoprover to prove all math by itself instantly.

Assume I write a tau or idris program that halts on the first zero of Riemann's Zeta which is not on the critical line. Of course, no one expect the prover to determine quickly if this program ever halt. If I knew how to do that magic, I'd already take $6M by solving the 6 math open problems of Clay institute Smiley

On the other hand, since the logic is decidable, the provable statements are exactly all correct statements and vice versa. Since every program has some runtime complexity, it is apriorily possible to give a proof for a runtime of any program.

So assume I wrote some simpler code with runtime complexity that I know how to compute and prove, then I can specify my proof or parts of it, and let the prover finish the missing details.

- Where can I find more (code / documents) about the JIT compilation of the Tau language?

This is exactly what I work on these days, so I'll have to finish first.

- Are there archives of IRC discussions / forum discussions / email discussions etc that can help understanding decisions made around the design?

I don't think there's anything one can easily track. ~10MB of text won't help anyone. So for now one can only ask and get answers, and with time we'll document everything.


Please let me know if I missed something or if other questions arise.

Tau-Chain & Agoras
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 27, 2015, 09:59:57 PM
 #247

Also I encourage you to play with http://attempto.ifi.uzh.ch/race/

Tau-Chain & Agoras
herkuljee
Jr. Member
*
Offline Offline

Activity: 34


View Profile
August 28, 2015, 06:13:11 AM
 #248




The marketcap is 7 million USD. The project looks cool but I'm not sure if this should be valued at more than 1 million dollars right now today.




This looks like a strong idea even though I don't understand much of the white paper details. Is Tau-Chain purely valued through this Agoras? If later someone wants to use Tau-Chain to solve some problem they buy Agoras (from market?) and pay for the usage of Tau-Chain network (how pricing the usage?). I understood inside Tau-Chain own rules can be set for that so can they make a rule to let someone use lot of computing power without charging anything? Or maybe I don't understand the logic where the value of Agoras might come..

If one Agoras is around 0.0006 and market cap is 7M USD or 30000 BTC (where to see the market cap?), then little over 50M Agoras is owned by people? And there is 147B Agoras to be shared. So 0.034% of the total amount is shared at the moment? Are my conclusion/numbers correct?
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 28, 2015, 06:33:45 AM
 #249

This looks like a strong idea even though I don't understand much of the white paper details. Is Tau-Chain purely valued through this Agoras? If later someone wants to use Tau-Chain to solve some problem they buy Agoras (from market?) and pay for the usage of Tau-Chain network (how pricing the usage?). I understood inside Tau-Chain own rules can be set for that so can they make a rule to let someone use lot of computing power without charging anything? Or maybe I don't understand the logic where the value of Agoras might come..

Tau is completely independent in Agoras or its coins. Agoras, on the other hand, is an application over tau. Everyone can write apps over tau - that's the very intention. So formally, tau as for itself is not funded. Also it does not come with a coin, but coins can be programmed over it.
Since after all one has to develop economical part over tau, using tau's unique features, we're going to do it as Agoras. And we do need to distribute the coins somehow, hence the presale.

If one Agoras is around 0.0006 and market cap is 7M USD or 30000 BTC (where to see the market cap?), then little over 50M Agoras is owned by people? And there is 147B Agoras to be shared. So 0.034% of the total amount is shared at the moment? Are my conclusion/numbers correct?

No, the 42M intermediate tokens represent 100% of all future coins, no matter if they'll be 147B or any other number, and the prices refer to those intermediate tokens.

Tau-Chain & Agoras
LOrrav
Newbie
*
Offline Offline

Activity: 26


View Profile
August 29, 2015, 12:39:44 AM
 #250


Since after all one has to develop economical part over tau, using tau's unique features, we're going to do it as Agoras. And we do need to distribute the coins somehow, hence the presale.


Can you explain more about this? What can I do, in terms of interfacing with Tau, with Agoras? Is it just another cryptocurrency that is just built using Tau, or does it have some special status or feature within Tau Chain? Why would anyone want to own Agoras over another cryptocurrency within Tau, or outside of Tau for that matter?
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 29, 2015, 01:03:25 AM
 #251


Since after all one has to develop economical part over tau, using tau's unique features, we're going to do it as Agoras. And we do need to distribute the coins somehow, hence the presale.

Can you explain more about this? What can I do, in terms of interfacing with Tau, with Agoras? Is it just another cryptocurrency that is just built using Tau, or does it have some special status or feature within Tau Chain? Why would anyone want to own Agoras over another cryptocurrency within Tau, or outside of Tau for that matter?

Agoras is indeed "just another app" over tau. This is important to bear in mind. Everyone can create more currencies over tau, and tau doesn't have a coin as for itself.

But Agoras is of course far beyond a cryptocurrency. It will have markets for trading code for money in a totally secure and trustless way (as in Code and Money blogpost), allow renting computational power with or without verification of execution, renting professionals in video conference and micropayments, smart contracts, pay-for-proof wheter it's math homework or a program to write, decentralized intelligent search engine, and even more.

Tau-Chain & Agoras
lordoliver
Legendary
*
Offline Offline

Activity: 1260

expect(brain).toHaveBeenUsed()


View Profile
August 29, 2015, 07:36:27 AM
 #252

Is it possible to write already apps on some alpha version? I would like to try out ;-)
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 29, 2015, 07:39:16 AM
 #253

Is it possible to write already apps on some alpha version? I would like to try out ;-)

For now you can do on tau what existing RDF reasoners would do (specifically Euler[EYE]) and at the same syntax, which is Turing complete btw. When we'll finish the typesystem it'll be essentially different, and ofc tau will cease to be Turing complete at that point.

Tau-Chain & Agoras
herkuljee
Jr. Member
*
Offline Offline

Activity: 34


View Profile
August 29, 2015, 11:49:46 AM
 #254


Since after all one has to develop economical part over tau, using tau's unique features, we're going to do it as Agoras. And we do need to distribute the coins somehow, hence the presale.

Can you explain more about this? What can I do, in terms of interfacing with Tau, with Agoras? Is it just another cryptocurrency that is just built using Tau, or does it have some special status or feature within Tau Chain? Why would anyone want to own Agoras over another cryptocurrency within Tau, or outside of Tau for that matter?

Agoras is indeed "just another app" over tau. This is important to bear in mind. Everyone can create more currencies over tau, and tau doesn't have a coin as for itself.

But Agoras is of course far beyond a cryptocurrency. It will have markets for trading code for money in a totally secure and trustless way (as in Code and Money blogpost), allow renting computational power with or without verification of execution, renting professionals in video conference and micropayments, smart contracts, pay-for-proof wheter it's math homework or a program to write, decentralized intelligent search engine, and even more.

What makes people who trade code asking Agoras in return of their services? Same applies for the other suggested markets, what obliges people to start using Agoras? They could ask, accept and take e.g. BTC, fiat or some other “app currency” instead?

Is there strong enough incentive for early adaptors and early coders to start asking compensations and stick in Agoras and what is that incentive? How sure it is that Agoras will be accepted as “the money” over tau? If not so and there’s no trading volume after IPO, price starts going down and coders easily abandon the Agoras ship…
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 29, 2015, 11:57:30 AM
 #255


Since after all one has to develop economical part over tau, using tau's unique features, we're going to do it as Agoras. And we do need to distribute the coins somehow, hence the presale.

Can you explain more about this? What can I do, in terms of interfacing with Tau, with Agoras? Is it just another cryptocurrency that is just built using Tau, or does it have some special status or feature within Tau Chain? Why would anyone want to own Agoras over another cryptocurrency within Tau, or outside of Tau for that matter?

Agoras is indeed "just another app" over tau. This is important to bear in mind. Everyone can create more currencies over tau, and tau doesn't have a coin as for itself.

But Agoras is of course far beyond a cryptocurrency. It will have markets for trading code for money in a totally secure and trustless way (as in Code and Money blogpost), allow renting computational power with or without verification of execution, renting professionals in video conference and micropayments, smart contracts, pay-for-proof wheter it's math homework or a program to write, decentralized intelligent search engine, and even more.

What makes people who trade code asking Agoras in return of their services? Same applies for the other suggested markets, what obliges people to start using Agoras? They could ask, accept and take e.g. BTC, fiat or some other “app currency” instead?

Is there strong enough incentive for early adaptors and early coders to start asking compensations and stick in Agoras and what is that incentive? How sure it is that Agoras will be accepted as “the money” over tau? If not so and there’s no trading volume after IPO, price starts going down and coders easily abandon the Agoras ship…

Accepting BTC or fiat for features elevating tau's unique properties (such as decidability and proofs over code) is of course impossible - such features are impossible to be implement over existing chains, even ethereum's.
As for the risk that someone else will write another and better Agoras, I think such risk exists in every project that comes to mind.

Tau-Chain & Agoras
freedomfighter
Full Member
***
Offline Offline

Activity: 210


View Profile
August 29, 2015, 05:58:03 PM
 #256


Since after all one has to develop economical part over tau, using tau's unique features, we're going to do it as Agoras. And we do need to distribute the coins somehow, hence the presale.

Can you explain more about this? What can I do, in terms of interfacing with Tau, with Agoras? Is it just another cryptocurrency that is just built using Tau, or does it have some special status or feature within Tau Chain? Why would anyone want to own Agoras over another cryptocurrency within Tau, or outside of Tau for that matter?

Agoras is indeed "just another app" over tau. This is important to bear in mind. Everyone can create more currencies over tau, and tau doesn't have a coin as for itself.

But Agoras is of course far beyond a cryptocurrency. It will have markets for trading code for money in a totally secure and trustless way (as in Code and Money blogpost), allow renting computational power with or without verification of execution, renting professionals in video conference and micropayments, smart contracts, pay-for-proof wheter it's math homework or a program to write, decentralized intelligent search engine, and even more.

What makes people who trade code asking Agoras in return of their services? Same applies for the other suggested markets, what obliges people to start using Agoras? They could ask, accept and take e.g. BTC, fiat or some other “app currency” instead?

Is there strong enough incentive for early adaptors and early coders to start asking compensations and stick in Agoras and what is that incentive? How sure it is that Agoras will be accepted as “the money” over tau? If not so and there’s no trading volume after IPO, price starts going down and coders easily abandon the Agoras ship…

Accepting BTC or fiat for features elevating tau's unique properties (such as decidability and proofs over code) is of course impossible - such features are impossible to be implement over existing chains, even ethereum's.
As for the risk that someone else will write another and better Agoras, I think such risk exists in every project that comes to mind.

Maybe to phrase what he is asking differently: after various markets are created over Tau (Agoras that will require the Agora coin to use them)-- what stops one to create competing markets, also over tau (parallel to Agoras) with BTC, ETC, or any other crypto tools. Won't these have an advantage over the Agoras market? (an advantage in the sense that they'll utilize other coins without limits).

OR, even if the above is correct, is your intention with the Agoras to enable features that can utilize any coin but that like in ETHEREUM, will require Agoras an an underlying gas/fee/whatever for every transaction while also enabling them as a "classic coin" within that Inner world?
klosure
Jr. Member
*
Offline Offline

Activity: 49


View Profile
August 29, 2015, 06:45:08 PM
 #257

- Do you have a specification (ontology, BNF grammar, spec document...) of the Tau language as it is now and/or as it is supposed to become once all the networking and crypto features are built in?
The syntax of the language itself is not new. The RDF family is basically all syntax-sugaring to NQuads http://www.w3.org/TR/n-quads/ so all RDF languages can be converted to NQuads, while plenty of converters out there. Current tau's code can read NQuads, JSON-LD, and (yet partial) Notation3.
I wasn't asking about RDF triples and NQuads: these are just a representation format. What I am asking is the specification of the Tau language itself. OWL2 for instance is the RDF implementation of a logic language called SROIQ(D) formally specified here that is a subset of description logic and its derivation as RDF based ontology is specified in great details here.
We know from the whitepaper that Tau is borrowing to constructive / intuitionist logic. But is it a subset, a variation, or does it implement the full logic? Is there a whitepaper or document covering the formalism? Is there a document that describes the ontology including the core logic and library of built-ins related properties?

Supporting network features etc. is done via builtins, which is also a common mechanism among RDF reasoners. See for example http://www.w3.org/2000/10/swap/doc/CwmBuiltins
Is Tau entirely structured around CWM? What about the builtins that are not already implemented by CWM? For instance DHT related features, RNG etc. Will you contribute them to the original CWM? Are you going to fork CWM to extend it? Or does CWM come with a way to register new builtins and if it does is it statical (compile time) or dynamic (runtime)?

- Where can I find a Tau code example that emulates simple procedural behavior like a loop for instance?

Let me use the cwm's builtins I just linked to demonstrate printing numbers from 1 to 10:
1 log:outputString 1.
{ ?x log:outputString ?x. (1 ?x) math:sum ?y. ?y math:lessThan 11 } => { ?y log:outputString ?y }.
see http://pastebin.com/H2gqLUzC
You are using the W3C swap/math and swap/log ontologies.
Are they a subset of the Tau ontology?
Again, that would need to be clarified. What IS valid Tau and what is not?

- How do you plan to make the transition from knowledge (decision of spawning a new socket for instance) to action (making a system call to create a new socket): will each i/o module poll the knowledge base periodically for relevant action? Will you place control points at key steps of the inference process to intercept propositions with a known action semantic and take action?
- Will input from i/o module trigger asynchronous reasonning or will it just add new propositions in the knowledge base for some synchronous process to reason with at next iteration?
At the syntax level, IO is just another builtin predicate, like { ?z a tau:tcpEndpoint } => { ?y tau:tcpConnect ?z }. At the typesystem level, we of course have Effect types/Monads https://en.wikipedia.org/wiki/Monad_(functional_programming)
The reasoning flow does not change at all due to IO (in general), but types originated from IO are tagged as effect types and of course aren't treated as "pure truth". A good reference is how Idris use side effects http://eb.host.cs.st-andrews.ac.uk/drafts/eff-tutorial.pdf
If I understand correctly your reply I/O originated events and data are being incorporated as new facts that will be (lazy-)evaluated / stated next time the reasoner does a resolution that needs i/o related facts, which in other words means synchronous i/o. I can see how this is going to add a lot of latency, but let me read that Idris paper first and I'll get back with more questions

- Is the interface between the knowledge base and the system level functionalities already developed or specified? Where can I find further details?
Apologies but I didn't really understand the question.
You gave this example above: "{ ?z a tau:tcpEndpoint } => { ?y tau:tcpConnect ?z }".
Are the built-ins behind tau:tcpEndpoint and tau:tcpConnect already developed or specified in a document?
What will be all the built-ins introduced by Tau?

- How can Tau guarantee that it can determine the upper bound of when a program can halt in less time than it would take to run the program? What if the program is already the shortest possible program to solve the problem it represents? At most I could imagine how code analysis could help determining a big-o estimate of the program complexity but then that won't help much to make sure that the program can run within the time boundaries of a given blockhain block as was suggested in the cointelegraph interview.

This goes right back to decidability, and to the point that one shouldn't expect the autoprover to prove all math by itself instantly.

Let me quote HMC's reply to the question "How does it compare to Ethereum" in the interview on CoinTelegraph
Quote from: HMC
To contrast, Tau is a decidable complete logic, per block, with Turing completeness recovered by the iteration of multiple blocks.  This means that we cannot compute any computable thing in one block, but we can know exactly how much resource expenditure a block will require to verify. We can still compute any computable thing; it may just require structuring the execution to run over the span of multiple blocks.  Because of this, Tau resources do not need to be backed directly by a token, and resource exchange can be negotiated by the users as they see fit.

In the context of the comparition with Ethereum, what I understand HMC to be saying is that the question of the maximum runtime of a contract is decidable which allows to break down the contract in execution segments and schedule them over multiple blocks. Hence my question. If the only thing  that is decidable for any program (you don't control what contract creators put in a contrat) is the fact that the program will stop, this does absolutely nothing to make it anywhere near practical. If the program takes 5 years to stop, we aren't any better off than if it didn't stop at all.

So if I understand correctly, what HMC says above would work only for these contracts that were shipped together with a proof already written explicitely by the contract developer that establishes in polynomial time that the program will stop because the developer went through the deliberate design decision of making his program do everything in small chopped batches with yielding and continuations all over the place. Is that the spirit? And to ensure that every contract on the chain are following this paradigm, the contracts would need to come with the proof for the transaction to be validated and added to the blockchain? And do you also ship a proof that the proof indeed runs in polynomial time? And so on Smiley?

Any example of a chopped contract?

Assume I write a tau or idris program that halts on the first zero of Riemann's Zeta which is not on the critical line. Of course, no one expect the prover to determine quickly if this program ever halt. If I knew how to do that magic, I'd already take $6M by solving the 6 math open problems of Clay institute Smiley
That's the point. So Tau will need to rely on a very self-constrained ultra-modular programming paradigm so that it is possible to prove that every chunk of code is both space and time bounded regardless of its input, while making sure that the proof itself doesn't become more complex than the chunk of code it vouches for...

I'm really curious how that will work in practice. A piece of code that demonstrates this would be more than welcome.

So assume I wrote some simpler code with runtime complexity that I know how to compute and prove, then I can specify my proof or parts of it, and let the prover finish the missing details.
Tau, the first Proof-of-PhD smart contract system Smiley

- Are there archives of IRC discussions / forum discussions / email discussions etc that can help understanding decisions made around the design?
I don't think there's anything one can easily track. ~10MB of text won't help anyone. So for now one can only ask and get answers, and with time we'll document everything.
I don't mind finding my way in a 10MB dump. I don't want to waste your time answering questions if the answers are already there somewhere.

Please let me know if I missed something or if other questions arise.
Proposed that nicely, how can I refuse? Here are a few more questions:
- Do you have specs about what exactly Agora is going to be. I mean other than the story of Bob and Alice . What kind of assets will Agora allow to trade? Will it features a full fledged distributed exchange like Ripple? Will it support other market paradigms like auctions or prediction markets?
- Have you started working on Agora?
- How long do you estimate you will need to complete Agora?
- Tau ETA?
- Agora ETA?
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 30, 2015, 03:08:16 AM
 #258

Maybe to phrase what he is asking differently: after various markets are created over Tau (Agoras that will require the Agora coin to use them)-- what stops one to create competing markets, also over tau (parallel to Agoras) with BTC, ETC, or any other crypto tools. Won't these have an advantage over the Agoras market? (an advantage in the sense that they'll utilize other coins without limits).

OR, even if the above is correct, is your intention with the Agoras to enable features that can utilize any coin but that like in ETHEREUM, will require Agoras an an underlying gas/fee/whatever for every transaction while also enabling them as a "classic coin" within that Inner world?

Thanks for the questions,

1. Whatever can be implemented with "old" coins like BTC, will probably be supported on Agoras, of course with limited features.
2. There will be no "gas" or fees for using tau or agoras. Otherwise, someone will create a clone with no fees, as you pointed out. We do try to come up with a design that is hard to compete with, of course, by all means we find.
3. No one essentialy prevents someone from creating Agoras competitor over tau, but it'll be hard to compete with tau's developers, especially in introducing the features we've been thinking on for over a year. No one even came up with a "small and simple" feature of agoras like Zennet. But again I think this points to a kind of risk that exists in any project, and usually doesn't come realistic, yet we do keep that in mind and try to make it more valuable to join us rather compete with us.

Please let me know if this answers.

Tau-Chain & Agoras
ohad
Hero Member
*****
Offline Offline

Activity: 878

http://idni.org


View Profile WWW
August 30, 2015, 03:08:35 AM
 #259

- Do you have a specification (ontology, BNF grammar, spec document...) of the Tau language as it is now and/or as it is supposed to become once all the networking and crypto features are built in?
The syntax of the language itself is not new. The RDF family is basically all syntax-sugaring to NQuads http://www.w3.org/TR/n-quads/ so all RDF languages can be converted to NQuads, while plenty of converters out there. Current tau's code can read NQuads, JSON-LD, and (yet partial) Notation3.
I wasn't asking about RDF triples and NQuads: these are just a representation format. What I am asking is the specification of the Tau language itself. OWL2 for instance is the RDF implementation of a logic language called SROIQ(D) formally specified here that is a subset of description logic and its derivation as RDF based ontology is specified in great details here.
We know from the whitepaper that Tau is borrowing to constructive / intuitionist logic. But is it a subset, a variation, or does it implement the full logic? Is there a whitepaper or document covering the formalism? Is there a document that describes the ontology including the core logic and library of built-ins related properties?

Supporting network features etc. is done via builtins, which is also a common mechanism among RDF reasoners. See for example http://www.w3.org/2000/10/swap/doc/CwmBuiltins
Is Tau entirely structured around CWM? What about the builtins that are not already implemented by CWM? For instance DHT related features, RNG etc. Will you contribute them to the original CWM? Are you going to fork CWM to extend it? Or does CWM come with a way to register new builtins and if it does is it statical (compile time) or dynamic (runtime)?

- Where can I find a Tau code example that emulates simple procedural behavior like a loop for instance?

Let me use the cwm's builtins I just linked to demonstrate printing numbers from 1 to 10:
1 log:outputString 1.
{ ?x log:outputString ?x. (1 ?x) math:sum ?y. ?y math:lessThan 11 } => { ?y log:outputString ?y }.
see http://pastebin.com/H2gqLUzC
You are using the W3C swap/math and swap/log ontologies.
Are they a subset of the Tau ontology?
Again, that would need to be clarified. What IS valid Tau and what is not?

- How do you plan to make the transition from knowledge (decision of spawning a new socket for instance) to action (making a system call to create a new socket): will each i/o module poll the knowledge base periodically for relevant action? Will you place control points at key steps of the inference process to intercept propositions with a known action semantic and take action?
- Will input from i/o module trigger asynchronous reasonning or will it just add new propositions in the knowledge base for some synchronous process to reason with at next iteration?
At the syntax level, IO is just another builtin predicate, like { ?z a tau:tcpEndpoint } => { ?y tau:tcpConnect ?z }. At the typesystem level, we of course have Effect types/Monads https://en.wikipedia.org/wiki/Monad_(functional_programming)
The reasoning flow does not change at all due to IO (in general), but types originated from IO are tagged as effect types and of course aren't treated as "pure truth". A good reference is how Idris use side effects http://eb.host.cs.st-andrews.ac.uk/drafts/eff-tutorial.pdf
If I understand correctly your reply I/O originated events and data are being incorporated as new facts that will be (lazy-)evaluated / stated next time the reasoner does a resolution that needs i/o related facts, which in other words means synchronous i/o. I can see how this is going to add a lot of latency, but let me read that Idris paper first and I'll get back with more questions

- Is the interface between the knowledge base and the system level functionalities already developed or specified? Where can I find further details?
Apologies but I didn't really understand the question.
You gave this example above: "{ ?z a tau:tcpEndpoint } => { ?y tau:tcpConnect ?z }".
Are the built-ins behind tau:tcpEndpoint and tau:tcpConnect already developed or specified in a document?
What will be all the built-ins introduced by Tau?

- How can Tau guarantee that it can determine the upper bound of when a program can halt in less time than it would take to run the program? What if the program is already the shortest possible program to solve the problem it represents? At most I could imagine how code analysis could help determining a big-o estimate of the program complexity but then that won't help much to make sure that the program can run within the time boundaries of a given blockhain block as was suggested in the cointelegraph interview.

This goes right back to decidability, and to the point that one shouldn't expect the autoprover to prove all math by itself instantly.

Let me quote HMC's reply to the question "How does it compare to Ethereum" in the interview on CoinTelegraph
Quote from: HMC
To contrast, Tau is a decidable complete logic, per block, with Turing completeness recovered by the iteration of multiple blocks.  This means that we cannot compute any computable thing in one block, but we can know exactly how much resource expenditure a block will require to verify. We can still compute any computable thing; it may just require structuring the execution to run over the span of multiple blocks.  Because of this, Tau resources do not need to be backed directly by a token, and resource exchange can be negotiated by the users as they see fit.

In the context of the comparition with Ethereum, what I understand HMC to be saying is that the question of the maximum runtime of a contract is decidable which allows to break down the contract in execution segments and schedule them over multiple blocks. Hence my question. If the only thing  that is decidable for any program (you don't control what contract creators put in a contrat) is the fact that the program will stop, this does absolutely nothing to make it anywhere near practical. If the program takes 5 years to stop, we aren't any better off than if it didn't stop at all.

So if I understand correctly, what HMC says above would work only for these contracts that were shipped together with a proof already written explicitely by the contract developer that establishes in polynomial time that the program will stop because the developer went through the deliberate design decision of making his program do everything in small chopped batches with yielding and continuations all over the place. Is that the spirit? And to ensure that every contract on the chain are following this paradigm, the contracts would need to come with the proof for the transaction to be validated and added to the blockchain? And do you also ship a proof that the proof indeed runs in polynomial time? And so on Smiley?

Any example of a chopped contract?

Assume I write a tau or idris program that halts on the first zero of Riemann's Zeta which is not on the critical line. Of course, no one expect the prover to determine quickly if this program ever halt. If I knew how to do that magic, I'd already take $6M by solving the 6 math open problems of Clay institute Smiley
That's the point. So Tau will need to rely on a very self-constrained ultra-modular programming paradigm so that it is possible to prove that every chunk of code is both space and time bounded regardless of its input, while making sure that the proof itself doesn't become more complex than the chunk of code it vouches for...

I'm really curious how that will work in practice. A piece of code that demonstrates this would be more than welcome.

So assume I wrote some simpler code with runtime complexity that I know how to compute and prove, then I can specify my proof or parts of it, and let the prover finish the missing details.
Tau, the first Proof-of-PhD smart contract system Smiley

- Are there archives of IRC discussions / forum discussions / email discussions etc that can help understanding decisions made around the design?
I don't think there's anything one can easily track. ~10MB of text won't help anyone. So for now one can only ask and get answers, and with time we'll document everything.
I don't mind finding my way in a 10MB dump. I don't want to waste your time answering questions if the answers are already there somewhere.

Please let me know if I missed something or if other questions arise.
Proposed that nicely, how can I refuse? Here are a few more questions:
- Do you have specs about what exactly Agora is going to be. I mean other than the story of Bob and Alice . What kind of assets will Agora allow to trade? Will it features a full fledged distributed exchange like Ripple? Will it support other market paradigms like auctions or prediction markets?
- Have you started working on Agora?
- How long do you estimate you will need to complete Agora?
- Tau ETA?
- Agora ETA?

Thanks for your questions, they raise important clarifications:

1. The OWL family is indeed a widely used ontology and known to suffer from logical complexities. Our goal is to implement RDFS (which is a relatively low level at this scope), and above it implement our type system (which probably amounts to adding predicates part-of and is-type-of, a good ref is ceur-ws.org/Vol-354/p63.pdf), and to recover OWL2 over those primitives.
I do not claim that any OWL ontology would be interpreted on tau just as any other OWL reasoner, and I do not promise any "easy migration" of existing code, as long as it is beyond just facts (i.e. without vars or rules).
So what I can tell now about tau's lang is:
a. Syntax is RDF family.
b. RDFS builtins will be supported as-is.
c. OWL will be supported as well but might not keep its common semantics' subtelties, as they change from reasoner to reasoner.
d. The rdf:Type predicate will be interpreted in a Martin-Lof dependent-type framework, as well as typechecking.
e. The language will also have Auth types for proof of execution (as in lambda-auth) and Effect types for IO.
f. The reasoner will be shipped with some builtins, this touches your next question and I'll emphasize on this in a moment.


2. The logic itself can describe everything that is computable. In other words, "all math that matters". It is so strong that it is considered as a new foundation of mathematics in general, see for example HoTT (Homotopy Type Theory). A great feature in Martin-Lof type theory is that you cannot even express a contradiction (like you can't even express Russel's paradox), therefore you never inconsistent.

3. Tau is not structured upon CWM at all. We do use Euler and Yieldprolog as references for inference and Idris/Agda as reference for typesystem, but not CWM. Moreover, I'm not aware of open-source stand-alone performance-oriented C++ reasoners out there that support N3Logic. We write everything from scratch, and in fact we already finished writing the part convering reasoning over triples. From here (except some more parsing features) is it all only features that RDF reasoners haven't met yet, at least not directly (like euler.yap jitting indirectly over swipl), to my knowledge.

4. What is valid Tau and what not is still early to answer. We will have to finish developing the language before we can say firmly "this is it".

5. To write "good" tau code one would definitely have to give a very good thinking, like any other construction. I do not claim that every first grader would write decentralized application over tau in natural language. But tau lets you focus on the logic and meaning of what you say, and help you express things unambiguously. Even if this will turn down to be more complex than say Python (which is of course a matter of taste), there are two important point to take into account:

a. Tau can have more frontends. The "real" interface is nothing but quads. Any sugaring over them can be de-sugared and fed to tau.
b. The more important point is code reuse, and to picture how programming would actually take place over tau. Say you want to write some complex program, and begin to break it up to functions and structures. Say for example that one required structure is red-black tree and one required function is balancing the tree. What you'd do nowadays is not reinvent the wheel but google this and find some code snippets for reuse. Those snippets may or may not fit your needs, and you still have to adapt them (if not reimplement them) to fit your code.
On the other hand, given tau, you can formalize only the requirements from the structure and the functions, and from here you have three options: either that code already exists in tau's records (might even be that your formal reqs themselves already exist) and you can find code that provably meet your needs, or, you can offer a reward to anyone that provable meets your requirements, or, you can write the code yourself, but this time the reasoner can give you a great help since it can prove you're correct or wrong in many cases. At the end, you end up with code that you can trust even more than after QA, since you have a mathematical proof that it meets its requirements.

6. It is worth mentioning that we intent to postpone to post-genesis everything that can be postponed to post-genesis. The a-f points I mentioned above are essential at-genesis, together with more features of course. Even the JIT is needed in order to specify the order of execution (possibly to future compilers) so client will be able to agree upon proof of execution. At this theme, all we do is give a reference, usable, and good enough client. Of course other implementations of the tau protocol are possible, just like there are many different bitcoin clients.
This might also shed some light on the common wondering "it must take years to build that monster". We indeed tend to pick the shortest path, and we believe it is shorter than estimated at first sight.

7. The IRC channel has become "home" for several friends and contain many personal and off-topic conversations. This is why I never published the logs in public. Nevertheless, you can simply join the channel and see right there how to get the logs right away.

8. I don't have specs for tau or agoras more than already published. Things are fluid during dev, ofc, and we better finish the client before we can give exact directions how to use it or what would be the features. Pretty much the same for ETA. I can throw numbers into air while apriori every number would be wrong. I can tell that 6 months till tau's genesis sounds horribly pessimistic for me, and another 6m months for agoras sounds strongly pessimistic. Sorry for being vague, but think of it for a second, and see that any other answer of mine would be a lie, and all I can do is give you the most truthful answers I have myself.

9. Have a look at http://www.cs.ru.nl/~freek/100 of course the syntax is terrible, and this, too, is why we build tau (HMC: apologies for using your catchphrase).

10. " If the only thing  that is decidable for any program (you don't control what contract creators put in a contrat) is the fact that the program will stop, this does absolutely nothing to make it anywhere near practical. If the program takes 5 years to stop, we aren't any better off than if it didn't stop at all." -- No, *everything* that can be expressed on the language is decidable! You can prove *all* *correct* theorems, and you can't even express their negation (i.e. expressing a contradiction).

I have to go now and I might add some more later. Will be happy to come back and see more questions Smiley

Tau-Chain & Agoras
klosure
Jr. Member
*
Offline Offline

Activity: 49


View Profile
August 30, 2015, 05:49:32 AM
 #260

Thanks for the answers and pointers.

From a stylistic perspective however I think the cross-sectional style in which you answer doesn't help clarity as there are bits of answers to every questions laid out in a completly different flow. Although this may seem to make more sense to you given that you have more perspective on how things fit together, people who are less familiar with the subject will have difficulties understanding how your answers address the questions. The numbering you introduce doesn't help since it doesn't map to anything in the set of questions.

Another problem of answering cross-sectionnally is that it lets a lot of stuff unanswered. For instance here you haven't replied about:
- the Tau specific builtins that you have already developed and/or specified.
- the discussion about HMC's quote and the way computation is meant to be broken down in code segments which time and space complexity is bounded and known (see the question again for the details) and the way continuation will be handled between these segments.
- how / whether you will have to prove that the proof itself can be verified in polynomial time and how to avoid this requirement leading to an infinite recursion of proofs.

Let's try to stick to the conventional quote-and-answer style going forward. Thanks.
Pages: « 1 2 3 4 5 6 7 8 9 10 11 12 [13] 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 ... 143 »
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!