Multi-chain Settlement with Anoma

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:08

   all right great this actually works let me get a Time on going as well all right perfect thank you very much today I'm going to talk a little bit about multichain settlement not necessar with nor more but rather to sort of in in specific content systems that we can deploy um in order to achieve mul chain moment um so with that let's get started so this is mostly talk on Typhon and Camara chains um and one sort of simple way to think of tyen and camar chains is that they generalize

2. 0:44

   shed sequences um a Shar sequen is generally something that can settle um across multiple systems that tend to have the same trust assumptions but Typhon on the other hand is designed for heterogeneous trust assumptions which pretty which is really the model that all chains are currently operating in um outside of sort of things built on top of e theum but so when you think of L Theory cosos these are all heterogeneous just assumptions and I mean I sort of uh spoke a bit

3. 1:14

   about this earlier on the panel I I think hetrogeneous trust asss is really what most people will want U because not everyone is going to want the same security model um most consensus implementations currently aren't aware of cross chain opportunity uh so like the consensus implementation has no idea what the state machine does with regards to cross chain messaging um Ty on the other hand here is actually

4. 1:39

   aware of cross chain opportunities and can take advantage of them um and this enables cross chain Atomic bundles with out rounds of synchronous messages uh between between uh transactions on different systems and I'll sort of go through a specific example in a little bit to make this a little bit clearer uh I think one thing we definitely looking for for collaboration on um which I think a lot of people are starting to think about is what does the API surface

5. 2:05

   both through user perspective solver perspective as well as a develop of perspective actually look like for cross chain um both applications of Worlds and Tents so if you're working on this kind of area um yeah please find me afterwards and maybe one other thing is the important thing here is the trust assumptions are not baked into the consensus protocol into typ orar CHS uh this can range impr proove of authority

6. 2:29

   to prove of work to prove of stake um to local government chain whatever you want this is completely trust assumption agnostic um so why does cross chain why do cross chain Atomic transactions actually matter well for one I think the future is going to be very much multichain as well as multitrust um seems unlikely that we all standardize on exactly the same chain with exactly

7. 2:53

   the same trust assumption for 7 billion people across the world this just seems like an unlikely outcome um so I will start with this very basic example of uh we have a train chain in green and a hotel chain in blue and the uh the thing that I want to achieve is I want to be able to book a train if and only if I also be in a hotel at the same time um so there are a couple of requirements here one uh we must Chu both transactions um on both chains right uh

8. 3:27

   this is sort of similar to the shared sequencer concept the system has been talking a bunch about um but this is actually not enough we don't just need that both chains include both transactions we also must have to succeed on both systems at the same time we can't have the green transaction failed transaction succeed right so we also need um something that umber research has been working on called

9. 3:53

   share sequencing um however sort of both shared sequencing as well as Shar abil sequencing it really works in the model of um uniform trust assumptions between the systems being connected here um yeah so um with camar with camar change withar petitions um we can actually allow for Atomic comms based on specific trust overlaps so imagine that you have a trust overlap between U blue and green

10. 4:25

    and blue and red but not red and green uh in this case we can form Atomic commits between the two things that have trust overlaps but not between the things that don't have trust overlaps unless you elect something else to provide uniform trust to both systems um one of the nice things is um every one of these chains can form effectively State petitions and you have two types of State petitions within each chain uh each chain Typhon has a base chain which really defines the quum um

11. 4:57

    that allows updates to the chain to happen uh think of this as the blockchain that defines stake as well as valid sets and rules on how to transition the blockchain for block by block right um and then we have chimer petitions which are where chains can move state that we want to get Atomic uh settlement with respect to other chains so for example here the red chain can form a

12. 5:21

    can have a Camaro State petition um that forms a effectively another chain in the middle here between red and that is atomic to both systems and we'll go over this what this means in a little bit so this is just the way the base chain looks um again this is a normal standard blockchain nothing very new here this is effective earning PBF or any other consensus that you like um okay now we've come to these C petitions so uh this is one chain this

13. 5:54

    is one blockchain um you have the base partition you have some State partitions or camar partitions um and these Camara petitions may contain out state for example they can contain things like a a specific hotel room that is bookable um the nice thing is it is relatively cheap to move State between the different uh State petitions on a single blockchain um and every state petition can read from every other state petitioner in the same chain um a number of blocks back so

14. 6:25

    you can read recent data uh which means you can actually very quick unlocks where um if for example this state petition gives up control over a specific hotel room the other state petition that may be trying to form a cha chain with a different blockchain U Can very quickly figure out oh the hotel room is now unlocked on this state petition it is now available over here and they can effectively take the lock there um yes this is how the starts look and I'm realizing the diagrams are very

15. 6:56

    small for the screen um with multiple chains so we again train Hotel plane um and this here on the hotel one is effectively the Camara partition that wants to have an atomic um commitment together with a train with a train partition um and yeah so you can only move State like you cannot move the same hotel room uh to both the petitions you can only move to one of them um thereby

16. 7:26

    you have no conflict between these and oh this and this um Y and really one of the simplest mental models you can form is that Typhon with Camara partitions enables on demand blockchains so this is sort of what we historically call a blockchain these are three different blockchains and whenever you want to do an atomic settlement between the two of between any number of them you form this Camara chain which is really you can view as a virtual

17. 7:58

    blockchain that just get spun up on demand in order to execute a specific cross chain settlement that you'd like yeah and so now we get to Atomic bundles how do these Atomic bundles actually look well anoma is all about intense and I think everyone is sort of still desar to realize that really everything in the future will be intense because they're by far the best abstraction for what users actually want to do with these systems so in this case

18. 8:25

    users will issue tents that say I want to um book a train as well as a motel um and I want to do so automically users will have to pay for this right like this is not free but users can decide how much they want to pay and if this is above the actual cost of executing this then um these chains can form um effective PS and then solvers um can form these Atomic bundles so this is a nice thing a user doesn't have to figure out how to interact with all these systems the user just has to in their

19. 8:56

    intent Define the settlement condition and the atomicity guarantees that they care about so then a solver sort of picks us up figures out how you want to combine these things and now you get to an atomic bundle which is being executed on this on demand chera chain if you have perfect overlap between security models of each chain um this looks very much like shed sequencing however you even get the atomicity guarantees as long as you have sound on um so this effectively

20. 9:28

    generalizes Shar sequencing to work in the model of complete trust overlap to no trust overlap and it rather allows users and ecosystem to sort of to figure out exactly what kind of overlap overlaps you want um and I'll will go a little bit Ah One important thing is I'm always using the example of like two chains for a Camara chain because it's visually easy uh you can have any number

21. 9:53

    of chains forming chains you can have like three or 15 different chains forming one and um Chimera chain onto which we can automatically settle things yeah and overall this allows for US Atomic cross chain communication without any um extra trust assumptions outside of what you have in terms of the unline security models of each involved train um and one of the really important things which are yeah I think most

22. 10:25

    people more people should be looking to database systems and database engineering um because effectively we just partitioning this database of each blockchain according to where it wants to Atomic or who it wants to be Atomic settlement with yes one of the really big benefits of this is that you can have that this is not a message based protocol this is not I'm sending a message over the wire I'm waiting for a bunch of consensus rounds to happen this is rather um this

23. 10:58

    Camara chain is by both validator sets and the validators have access to the entire state they're just running these transaction they're just running um transactions against primar chain blocks which means we can have synchronous message passing between these um two blocks right so like I can have a book me a train transaction over here that um within the protocol actually checks

24. 11:24

    whether the hotel room is available and if it's available um this returns to and then it execute so you can have these um very Atomic synchronized message passing with each other which means that you don't have to rely on Rel layers or solvers to actually do this kind of work for you rather you can have like a model that is very similar to how you can program existing synchronous systems like

25. 11:51

    media yeah and this is then how this starts looking if you have multiple ones of these at the same time um and and really the really nice component is that you can always read the state from the other petitions not just the base State petition but also from the camar state petitions um yeah so on to atomicity guarantees and sort of one some of the benefits here of Typhon um the higher the overlap between the quorums of each

26. 12:21

    involved chain uh the stronger the Miss guarantees um really if you have 100% overlap um your guarant is is talk no foring rules on the involved systems um so for example in the case of bft this is generally 2/3 plus um these optimistic guarantees make Ty and behave a little bit like a bridge except that we don't have to have an external Bridge provider operator that provides this uh consensus finality but rather this atomicity comes from the

27. 12:54

    consensus of each system jointly um and even if we lose atomicity guarantees because some people are misbehaving or some people are not well connected to the system we don't lose safety uh this is uh the worst thing that can happen is that the different chains uh order the transactions differently so on one chain the hotel room was booked on the other end wasn't but it's not that we can double book a hotel room or that we can

28. 13:20

    lose funds here um so yeah this is really important state is never lost nor Frozen nor corrupted yeah this is what actually ends up looking like if toiss is lost um the camar chain just diverged we didn't violate any of the rules of either base chain of either chain uh we just happen to have a different view of ordering of these resources this is not great but this is not um like a critical failure effectively um and the other big benefit is that in

29. 13:52

    order to get these atomicity guarantees we don't have to do these multiphase commits or um sort of go through multiple consensus right like when you think of IBC IBC you always have to move assets first lock them on some system and then do actions on them here you don't you don't have to move the assets um you also don't have to move an asset lock it free it for a while um and until you have all the Assets pren in place

30. 14:15

    and then you act on this um so you never lock assets unnecessary um yeah so some future areas of of research and collaborations that we quite interested in um so exploring the continuing between lifeless and safety um for example you can not just take the overlap of both chain both chains for example if you think of Val say you don't have to just take the overlap between them you could also take the union so that um this impacts you liess

31. 14:45

    a little bit because now um if more people go offline you lose lifeless earlier however it also increases the um safety guarantees to some extent to not lose atomicity um one interesting area is also you can use it totally different Corum to connect two dis joint systems so for example when you think of how you want to connect for example ni and salana um if they have no overlap in validator set you could choose to use

32. 15:10

    ethereum as the common thing um to run the Camara chain in order to get aity guarantees across both there Atomic settlement across between me and salana without them having to agree on a specific on running the same security model um yeah the other area is intend layouts and formats um things like encoding formance fense I think as an industry we just benefit from standardization here one thing to keep in mind they should be circuit friendly um the other thing is apis for solvers how do solvers actually interact all

33. 15:41

    these systems um yeah and then application apis I mean one of the sort of complex things here is what is the CIS called interface that allows uh chain a to call a transaction on chain a to call a transaction on chain B in the camar chain in that how to express different settlements um yeah there are some resources and Conta uh contacts um I would have you recommend all of these also check out JY Thea blog blog. an.net if you want to learn more um also we on

34. 16:14

    Twitter I'm on Twitter on telegram come find me there if you want or talk to me afterwards thank you very much