All Core Devs - Execution #69

Transcript

[Music] [Music] [Music]

[Music] [Music] [Music] [Applause]

[Music] [Music]

[Music] [Music] [Music]

[Music] [Music] [Music] you you

yeah I don't know what you mean by changes there's nothing where I was like Nicole I thought and most of the chatters getting one what what it really is when someone thought it was something else that okay I'm getting yep yep you're right I was reading Thomas's chat about the final accepted wording on 2200 and I thought that was a that was a thing but no you're right it's it's just clarifying questions so let's see yeah that's right yeah that's been right I need to add that you're the

code side by side so I actually found two flaws in Pantheon which I think they've fixed by now one of them being that for 1884 they when they implemented self balance they tried to do a caching solution so they cashed the balance at the beginning of the call frame and then just deliver that whenever self balance is invoked which doesn't work nice I just said that all right

Hudson Jameson: I just updated the agenda if y'all want to refresh your page and agenda item one

now has a link to Martin's tracker so that's really helpful looks like the Geth team has everything Peter or Martin do you want to elaborate on that at all

not much so we finished finishing plant a VIPs and merge them in we also enabled I mean implemented Istanbul configuration configurability from the Genesis back plus like to override it so anyone can play around with it would be nice to put together maybe a multi-client thing they just play around with it but from our perspective we just do

waiting for other people now so as of today we merge in support in the guest retesting Department for Istanbul we merge and support so you can run also the EVM with Istanbul Genesis support I mean the raw bare-bones EVM yeah yeah we also merge steamed in the postpone the override flag which yeah I'm hoping that the clients so in the event that we want to postpone it we can do so or rather users can do so from the command line

without downloading a new primary

Hudson Jameson: nice the next client is Aleph and the chart that Martin has is accurate as of 24 minutes ago when when gumbo on the agenda left a comment about the progress update for

a left and it looks like

Hudson Jameson: the next client would be parity

code it'll call data

kescott reduction merged and we have all the other except the Yankees our zoo for us to an MP are sore haven't be implemented so I've mentioned in in the theater we probably need to six of September to finish the implementation it's not only because regards exactly IP released late but also right now we are in we just happened to be very large

code base refactoring so we try to have some really large refactoring PR opened in our code days and we probably want to merge them first before emerged example yeah QPR's alright

Hudson Jameson: understood we already talked about Pantheon and they are reflected in the chart

Trinity

Hudson Jameson: as far as I can tell they still have the Blake F e IP potentially unimplemented as the call

data reduction unimplemented everything else looks either open or merged

let's see nevermind Oh everything accent blank is merged already and the blank one we're in the process and we've already prepared it recompile and we found a library with connecting library just need to make changes to extract this compression function so should be all soon ready and I'm ready to test

Hudson Jameson: great am I missing any clients

okay the next thing I want to discuss are the recent benchmarking also so there's one from Martin and there's one from axe ik I'll post the link to

axis xx's benchmarking right now no yell if it's Berlin let's just skip it but come back to it when we're talking more about the Berlin stuff and

Hudson Jameson: then it looks like there was some gas

estimation made for the verification of Z cache headers by AXA kits in a comment I'll paste it in chat and it has a link to let me see this just happens so I haven't read it at all oh it's okay it's in a comment so let's see here and short the call cost to compile 385,000 gas alone not counting anything else I could imagine the total cost would be something like 600,000 to 1 million gas so I think that's just some interesting

information about the Blake to be e IP does anyone else have comments on that

who might have read that or like understand more about what it means well I mean it's I think so

Martin Holst Swende: one thing that Isaac has proposed earlier is to reduce the gas costs for pre compiles and in relation to that EEP which I don't recall the number on this is of course interesting because if we reduce this number could go

like

an order-of-magnitude

Hudson Jameson: hmm anybody else have comments okay next we have decided the block number for Istanbul test net fork it sounds like based on a few clients that still have some implementation to do that we might not be able to make that decision today but I want to hear everybody's opinions on that and see if this is something that we want to do

today or not because

I'm open to hearing others opinions on that so yeah what did what is your estimates yeah we probably need like two weeks from six Osterman and that's for that's for your your release to be out right so we probably need some buffer after the sixth of September then hard for the client the

test nets correct yeah that's just a save date for us like if we like if we wash with my yeah

Tim Beiko: yeah that's the date just by which you feel comfortable having implemented the eeap's not the one by which you feel comfortable

hard-working the clients hard-working the test net yeah it's a rather comfortable having overseen see implemented

Hudson Jameson: got it and what's the release cycle like at parity is it like just when you want to or is it on like a time schedule

we can we can do like like we can do release like just after implemented all

the tribes I think that should be alright but probably like if we can have like I'd love at least a few days or a week buffer okay yeah I

Hudson Jameson: I think that if we have a short meeting next week or we decide on getter what the number should be I think doing it in a week would be safe because that gives you all you would have a better idea of how much more time you need and if it's anything more than a week we can reevaluate but if it's just a week then we can pick the block number and start having clients release is that

something that we can agree on everyone or does that sound a little too aggressive

Tim Beiko: so I think it's just worth highlighting that if we do this say that brings us to Argos 30th where we pick a block it'll be a week or two I assume after parody's release so mid-september ish and and it seems very well it just seems impossible then that we hit the October 4th main that deadline like two weeks of test net before main that seems a bit reckless so it's worth probably just making explicit that this also means

the main that hard work pay for

will slip unless there's something I'm missing yeah I just yeah like I just have one comment tonight date and it's not related to implementation specifications to gas turbines thing so I still think we might just want to fix that issue before we actually hard for the titanite because otherwise if we decide to do something else for eight eight four then like we cannot infer things and it may

create a lot more complication and actually delay them and that further I don't understand why what do you mean we should solve right now when it's like for the first guy Steve and I think we might want to like like Rice's so that we don't accidentally free some contracts right so why is it for yuppie write it for ya okay just - right now but you're

Martin Holst Swende: so you

would propose that we do that fun in combination with the general racing of the gas - from 2300 is something else

yes oh well so I think like if I'm understanding this correctly like like like you have done some analysis that we probably will freeze some contracts if we just apply it without a chant - gasps diamond so I think we probably want to do something just before we hard for tonight so that like we'd like so that we don't like need to like to do some

really complicated things in the future

Hudson Jameson: okay I think I know which is she you're talking about you're talking about the one where when we repriced the opcode from 1884 then it might freeze up some contracts that are dependent on the gas cost being a certain amount and their contracts I get that right so far

Wei Tang: yeah

Hudson Jameson: and so what happens is we want to prevent that sort of do we don't have to do anything tricky in the future to undo that mistake for stuck contracts and the analysis has been performed by Martin and they did find some contracts that

that would potentially break and I forgot what the analysis was exactly Martin can you refresh my memory

yes

Martin Holst Swende: so there is an analysis on a github repo mine and it spits hard it's Johanna sweeping general conclusions from it because in some cases the problem is that you have for example designated Sanders that are the only only only a few a couple of Sanders are allowed to deposit to a certain contract and those Sanders use the solidity mechanism transfer which only sounds 2300 and with

that so therefore the Sanders can only sell with 300 and the recipients does assload a notable example that was khyber so in that particular case an increase gas type stipend would work however there are other similar cases where at fixed amount of gas is used like nineteen thousand four hundred ninety thousand is used or some kind of similar call and it's kind of hard to and and other potential things that could break is if

a contract in a default function does does a few log operations and also an S load and there could be different ways to solve these problems and I don't really think that there is one solution that would solve it for every case and I don't even think that every case needs to be solved some contracts are dead and unused for like five hundred days other contracts are still used but they have built-in capabilities to upgrade themselves some kind of yeah a great path which I mean it would be more work

for the operators but it's not impossible so given that there's kind of no really self-evidence magic bullet to solve it's my personal thoughts is that which you just kind of go ahead and if there is something which we up then that could be fixed in the heart for following Istanbul and at that point in time the operators of

contracts who have been destroyed by this they can contact their old core devs and they can it makes it easier to figure out what the best solution is that is sufficiently yeah solves solves the actual problematic ones and not every theoretical corner case that exists on the blockchain

what I mean

Hudson Jameson: okay yeah I think I got that basically if we were to find a solution today we may find out after the fact that it was

not the best solution based on the feedback we get from the people with broken contracts is what I what I kind of interpreted from that

Péter Szilágyi: not exactly so we so there was a proposed solution but it's really really happy and weird and our suggestion was that it feels weird to add a hacky solution for for a case that might not be relevant and the idea was that let's not add the hack in now but we know what the heck could be and then if if we legitimately break something that we didn't know that we're going to break then we can always put in the hack later but let's not hack it before it's actually needed

well

Hudson Jameson: don't

we know it's going to break some though or do you mean break some and people complain

Martin Holst Swende: so yeah we know it's gonna break they're really good big things we don't know and all of those things can break in different ways some of them can break because of a combination with us load and load those could be sold by making log cheaper some other break because they just do a lot of s load in the e at the recipients and those could in some cases be solved just by upgrading the contracts blah blah blah so there are different ways to solve it and there's not really one the hacky solution would kind of open us up for a

tax optimal either and waste proposal but he mentioned earlier here to raise the gas type and it's it's kind of a can of worms in itself because that brings us back to issues with

reentrant C and state publications so that's yeah I seem analyzed yeah but like I think which number is it to fix for one to h3 or should already cover most cases for the obstacles rising writing the cash

stipend

Wei Tang: what I really worry is that is that this could be a potential PR disaster so the issue is basically if we if we broke some contracts and same like some contracts and we were asked asked who like unfreeze it and we could have a similar situation like the party Murray State so it's like all about subsidy contracts and our boats to make sure some contracts are unfrozen so it could be like a potential really policy

closing and I just worry in if we if we don't fix them right now like in the future we may may not have a chance for them to our function at all

Hudson Jameson: hmm that's not a bad point I would say that part of what can prevent that is kind of getting ahead of the kind of getting ahead of any negative PR that happens by completely explaining the

situation in ways everyone can understand and how this would be a different case in a way that these the way the contracts broke this time was not what some would consider an idiot sequency with solidity but a side effect of a needed upgrade that happened at the protocol level and not a higher level i just how I would explain it but I might be getting that wrong and I would absolutely consult with a lot of people before pulling something like that out there

like my mom who is I don't say like I still think it's a simple

operation and we should have most reentry possibility to the cupboard bite if you guys to 200 in so I just I just think I still don't think why rising Jocasta pan would be a big issue so there's obvious solutions like

Martin Holst Swende: the thing is we're currently there are theoretically up to 20 s loads you can do at 23 on the gas and if we want that still to be theoretically possible after we reach the 800 that means the gas stipend needs to pump up to like 16,000

something like that and if you have the gas type in at 16,000 then obviously you can do

reentrant see and state publications so it's still going to be like where do we put it and but but but

Wei Tang: you can't do reentry right if if Annie Annie I store is below the cash stipend and range would automatically fail after

her to time work

Martin Holst Swende: yeah we'd have to modify to to 200 as well and then all of a sudden you would be unable to do state modifications if

you have less than like 16,000 gas but it's it will be better if you so there have been a couple suggestions to how to handle this it will be better if you write down a proposal where you specify some exact numbers I know the contract library guys have volunteered that if we want to play with some scenarios such as no increase in the gas type and so on or lowering the load costs for another then they can evaluate that against set of affected contracts so we could work with them to to do some automated analysis of scenarios but I I am not seeing a

definite proposal to how we can solve

everything Martine something maybe

Tomasz Stańczak: controversial but of you thinking about it for a moment and I think that we can introduce something like a transaction counter of assload calls and only invoke the race of the s load called cost after now sorry it wouldn't work like we could potentially make a progressive cost but like not too complex but some kind of progressive cost for those operations

yeah we could either discuss it on

alternatively if people file tickets on the this security ke 1884 security con he painted for security repository fun tickets with proposal forever to fix it I don't know what's the best form

Hudson Jameson: yeah I'd say a combination of the awkward ebbs get our channel announcing the analysis and commenting on I mean I guess just cover all bases fellowship of a theory of magicians as well as trying to merge information and these security considerations of the EIP is that the third one you mentioned Martin security

considerations or is there a whole nother repo oh that's right so yeah if we could basically store the analysis on in the repo and then have a discussion on the chat and fellowship of etherium magicians that seems sufficient for getting the word out amongst the core developers for any proposals for something like this but I I agree with Martin that there needs to be something a little bit more like formal proposed about what's gonna be the best way if we're going to put some solution in

before the hard fork what that solution would be are there any other comments on this anybody have an idea or stuff like that

James Hancock: my comment is that this sort of this feels like a unique circumstance because it's kind of one of the first times this has happened at least the idea of an intentional change that does have some breaking changes it also is kind of the problem the promise of what each one acts needed to be or the observation so while this one case I may be thinking about it from solving this exact case as we keep moving forward this will happen more

especially if we get into state rent and so at the general case of how to approach this coming from this specific one I think it's worth thinking about

Wei Tang: I just want to comment there's actually a really really easy way to to fix like fix those things once forever with a breaking change so we can see see just an idea so we can just do a conversion in of version 0 and then for version 1 we remove any of codes related to cash so so we make it so that a contractor helper cannot make any assumption about

the best console so all those gas issues will bill just not happens forever and from that version where we can change the gas cost whatever one for what how many times we want to do future boots

or whatever

Martin Holst Swende: yeah but I don't see that being a solution because we still have the non version things right yeah we with at low gas so it doesn't solve the problem

yeah right so there we Vanilli rate with something like charging extra for version zero versus version one and my point

James Hancock: was even outside of gas costs like

there's a very large network of soon anonymous permissionless deployed who knows what like there isn't a hey this contract is owned by this person male 2 so just how to handle this situation where it's something we know could happen how do we proactively say hey we're thinking about it with the realization that it's kind of an impossible to reach everyone in the network so someone will likely get upset

so they'd say something oh you're muted James now you somehow clicked me oh yeah because I

was done oh that was it okay sorry it sounded like a cutoff mid word but I think yeah it was just an early yeah so so

Péter Szilágyi: I mean realistically if theorem is going to be alive for the for the coming however many years then we cannot expect never ever to break anything so I mean it's not I don't think it will ever be possible to keep indefinitely running contracts that were deployed in frontier I mean it would be nice of course and we should definitely strive for it but but if we if we say that we won't ever ever

ever break anything and that might actually be quite quite a heavy price to pay

Wei Tang: actually I think breaking stuff it's just that EVM is not really designed for future app wish but you can really make some simple change like I mentioned before just removing all the reference to gas to make EVM like like good for future up with friendlier to feature a please and and there's a like you can make this so that it's really hard to actually break stuff like too much a

really good

back workability well yeah

Péter Szilágyi: but you're talking about gas price changes now so it specifically nobody thought that this would be an issue and that's why nobody fixed it yes we can now create the solution to fix the gas prices issues but I'm almost certain that next time there will be a completely different class of issues that again would need something special to fix it so I think it's it's part of social development that you discover problems as you go along and you need to figure out what to do with those problems but also like white proves or theoretically prove that you know you're doing actually Greek stuff

you know I need friends yeah of course but what I am saying is that if are you willing to pay the price of not breaking anything

like if if we if there's a way to do that I would I would definitely do that because I think like just excessive one for for Frisian right because code is law like you don't want to break any weight I mean

Martin Holst Swende: we did back in the shanghai attacks we intentionally broke the attackers contracts that's what we did in the tangerine whistle its various

dragon we intentionally broke those contracts and the probably I mean do we want to wait until the next round of Shanghai attacks which targets are cheap ass loads before we do another upgrade of that and and at that time intentionally break those attacking contracts

I mean there's a there's a major difference between some attacking the network and we just fix it than having a future up ways that that brings some not attacking contracts I think those are really two different things

and also if you mentioned attacking contracts like like like it still the issue of Casco so you can you can actually fix it in different ways that doesn't break those attacking contracts for example makes a contractor expensive to run so so it cannot be used as attacking contracts anymore or stuff like that but as using like bike work availability is it was quite important soon for easier that I like I think we should really carefully but but for this Hartford I

mean this is just my like I'm just saying my opinion but the issue is a 8 a 4 is not like over my dead body that if we go with it I and Families it I think

James Hancock: I think it's also important to consider the other side not just where a change couldn't break a contract but where I change would be benefit benefit to another contract right for example the reducing of gas costs Mike Hayden came in today Al Gore Davis and said this is something I'd be great for you swap redeploying based on versioning would be really would be really difficult for them so there are there are people who would benefit from having a breaking change happen and there are people who would who would detriment from having I

could change this happen University are not versions and I think just what we have to swallow both pills if we want to do one

Hudson Jameson: I guess James the only thing is it's not a it's not a thing where we're swallowing both pills because we're what what ways proposing is putting in something that saves the old contracts while reducing the cost so that's kind of a an idea for something that will help with the PR and not make people upset and keep aetherium contracts in a state where at least for the moment we

can say the code will for the moment run as intended without breaking a contract except for all the other times we've done it

the ways idea has to be implemented in addition to 1884 right

it's it's not like just we can clarify I don't think we can get what I said in Eastham board so it would require a conversion and you require another lab for further subversion of and breaking change so I thought I just don't think we can get it for example but that's the thing I think we'll fix fix those kind of issues once and for all

okay well ice it sounds like yeah for this

Hudson Jameson: so you're saying for this issue we wouldn't be able to cut it for Istanbul so we have a good amount of time for any proposals that come out about ways to fix this if we decide to fix this at all is that accurate

Wei Tang: yeah like what I'm just seeing my concerns that I think the current situation should be treated more seriously like we need to actually make a decision by the rebound just break those contracts or or still want to fix it like this must like I think we we probably want to get more

people like outside of the core

- like from the IP well first whether they are

correct me if I'm wrong but it's not it's not like a nice to have it's more of an existential the chain will have serious problems if this if this gas costs doesn't change right or it could in the future

I don't think

Wei Tang: I don't think this is like accidental like like this ice law change from 200 to 800 is is just nice to have right this time

Martin Holst Swende: no I also consider it not a nice to have

but more important than that

can you expand and like why it's a necessary thing Martin

oh yeah

Hudson Jameson: it won't change the decision but yeah I'd like to hear why it's more necessary than a nice-to-have for if like for the reason of if we discuss this with the community how we can frame it

right because these people are gonna they're just looking at it from their baseline like oh now it's harder to do this thing that I'm doing but in reality it's not like it's not a punitive measure we're not trying to make things more difficult and they need to understand that it's it's like it has

Martin Holst Swende: yeah it's it's simple the case that does the state grows

larger and larger it becomes more io intense to look up something a random key in our state right so basically an attacking contract

which just spurious Li loads random slots will be quite expensive you know when I stick sconce I mean it would be it would lead to habit blocks yeah that's well I say me the future is to like when we have larger state which you need to rise ice Lord

again right so I think we might like we might like like what I'm saying we might just want to fix it's like find some way to fix it before deciding the test pad dates because so we yeah sorry I didn't mean to interrupt you know I was just saying like we might just want to find some way to fix it before deciding the test night a big house like we will still have the same usual next time because when the stage continued to blow it with you need to rise i snort

cashcourse again

Martin Holst Swende: so i consider this to be a band-aid until we get some proper 1x changes that can reduce the state load I'm hoping that we don't have to raise it again but instead have some benefits from phonics that means we don't have I totally agree that it would be really nice if we have some really good mechanism that stops the contracts from breaking I have not seen any such

solution that I think is the most ultimate of all of them there are three or four different things that could be done so let's continue that discussion and hope we get to something otherwise I would be fine with just trying to solve it's post facto

Hudson Jameson: we don't have to so we already did we are there's already the reality that this has to be solved after Istanbul as far as implementing it so we can discuss it between now and then as this is a top like an important topic of conversation however a formal proposal will need to be written in

order to have a baseline for a fix that is you know that was something we could actually do that might be a little less hacky or something that it would be a little more clear is that is that somewhat what you were saying Martin that we don't need to focus on this the second because it's not going to go into Istanbul anyway

no like several weeks focus on these who will say whether they

can come up with something good before second the test enacted just just so we'd like like because once we have this test nights it's really hard to for other solution might be more elegant to be applied on our test net like I'm like for this list to clarify I don't mean to delay my night or something so so this is just purely for the test night date and and I just think if we if we like don't play I would rather be safe than sorry for the

test night because if the test night actually has some other issues or got delays and the magnet will be delayed further

Martin Holst Swende: on the other hand the test net is pretty good it's actually pretty good to roll this out on test nuts and see what breaks and I think obviously a lot of crypto reads Twitter but those who don't might find on the test nuts that they are contracts broke which I'd argue is a pretty good data point

actually but

Wei Tang: but we already

know loopy it will break some contracts

so yeah but

Martin Holst Swende: we don't know what contracts we don't know which ones are still

relevant yeah like like what emotion is that

Wei Tang: it might be a PR issue like if we have like a list profound like like a test plan that is entirely new not a loop stand perfectly fun to to high-lows actually testing bad but there are a lot of people using the ropes and test night so if we accidentally break a lot of

contracts

like like like it will lower the confidence for people who look forward to his temperament night

Hudson Jameson: I think that from a PR perspective the core developers should worry less about that I mean that's obviously something that you know keep in mind but there are people who are communications and PR people within most of our organizations including parity and the etherium foundation and Pantheon and everywhere that can get together and craft a message that is very clear about what the mandate for one X was and how

these things are gonna go down now and in the future so that the core devs don't have to worry about that and some of it can be deflected I know I personally am probably going to write something to the extent of you know there is a communications breakdown between major DAP developers and the core devs and part of that is not having a position technically that's a paid position for someone to do that full-time and to find the best solutions for that so I guess what I'm saying is there is the potential for some bad PR but I don't think it's a show stopper for implementing something that has to

save all the contracts and I think there are people equipped to handle that and I I know the people from the EF parity and Pantheon who could get together and I think we're already in a chat room who would who would help with that so I I think that should be less of a concern honestly

Martin Holst Swende: yeah I like to stress but the counter opinion that to ways that actually since we probably will break some contract flows scenarios it's important to roll it out on test nuts as early as possible so they can

investigate how to remediate that situation and check if their upgrades are working or how otherwise they can fix

that's my

Hudson Jameson: hmm yeah I can see that perspective are there any final thoughts on this what it sounds like the conclusion has come to is we'll need something formal so we can discuss it on a technical level what would be the benefits of it and then we would need to describe on a more abstract level what we want the future narrative and direction of narratives

actually the wrong word the future direction of what these breaking changes are gonna mean and how we should best handle that with the consideration that these DAPs some of them being very important to the community could break and making sure that we keep an ear open to people who have those concerns within the community that's somewhat what I got as a conclusion does anyone have anything else to add to that that I kind of didn't go over as far as next steps okay thanks everybody this was I thought

this was a really good conversation to have so that we can kind of have expectations set for future upgrades that happen even post

a Stan bull

Hudson Jameson: do we have anyone that has an update on testing I know that there was some there was someone who mentioned that we should definitely have cross client testing be one of the priorities I forgot who mentioned that but is that something that is that something that anyone wants to comment about

so Dimitri called on the call

Donal has done some work done or IANA call yeah let me find my undo button

Danno Ferrin: yeah I've done some we're putting together some unit tests using the rate retested framework reference tests not just unit tests to try and get some basis to do some cross client testing the current retest stuff it does you know what kind of times you just have to line them up and do them you know one at a time currently I'm debugging a difference gets like two F is getting a different state route and Pantheon but with the tool we can get to the root cause of it I've written some tests up for some

basic self balance chain IDE and Lake to F and I was gonna be working on getting some of the other tests out there to make sure that they work as expected currently they're sitting in PRS as part of the etherium test frameworks they haven't been merged in the main yet I think before we merge a test in the main I think we should have at least two clients who agree on the results

Martin Holst Swende: yeah I don't think we need to be that careful about the merging because the clients need to I mean I didn't get something apart as well we from time to time

pulling the new tests it's not gonna optimize the thing

okay

Danno Ferrin: same with the Pantheon but even when it's called a mean validation I think from least two clients at some point that yeah this is the right number not something I made up

yeah I mean the fact that some at some point we're gonna have to or over clients need to that's like before we go live I know what's the status of hive I know Pantheon snot on hive yet I'm announcing

Martin Holst Swende: a major refactoring and yeah it's gonna be a big change to where

hopefully when in the next like week or two we're gonna spin up a new machine to put the new version of hive on and then transition from the one to the new one and yeah obvious it's a priority to have done when the tests the Istanbul tests are eventually done and exported into blockchain tests because that's that's when I start testing

okay so I just do the differential no and I cannot we cannot start until the client this case

parting gift actually implement Istanbul and make

Hudson Jameson: cool any other testing updates from anybody

yeah by the way there are transaction for the I was just wondering curious if

anyone was actually run them

Danno Ferrin: I'm not sure if we implemented the yep I haven't been able to dig into the co2 generated we have stuff referring the transactions tasks I tried pulling him in and nothing broke but that doesn't mean that I'm sure they were writing to me and I still need to dig in deeper there

yep cool

Hudson Jameson: anybody else have any comments

okay

Hudson Jameson: the last thing we'll need to do is kind of go over some of the

action items all the decisions from last all the decisions from last meeting are very straightforward so I'm not going to go over them the only one that's a decision that is now being changed is that all clients are required to have implemented all the IPS for the hard fork Istanbul won by the 23rd of August 2019 that's been pushed back to the 6th of September and the decision for the block number did not happen today so we're still discussing that and we can also continue to discuss that via Geter so those are the two decisions from last week that have just been changed otherwise all the decisions are pretty

forward

oh yeah on that

Tim Beiko: given that we're pushing back it's probably worth making it not sure if it shows the right way but should say that October 4th may net hard Fork is probably not going to happen so I just want the double check that with everyone here because if people are expecting yeah sort of made that hard fork before DEFCON it just seems impossible at this point

Hudson Jameson: I can agree with that does anyone else have any comments on the possibility of that Oh Daniel you're unmuted sorry so who

thinks that we're gonna have to pick another date for the main that hard fork

Péter Szilágyi: yes I think that's pretty obvious so honestly I would say at least one month after the

test net is forked so that's kind of what I was thinking - third week in November and then we moving out again if we know we're gonna miss it

Hudson Jameson: let's not put a date yet exactly on it until we get the

test net number I'll go ahead Peter I think I interrupted you know I wanted to say the same thing

that's wait for a test number and I'm cool

Hudson Jameson: anybody else have comments and on the topic of PR for those from Coyne desk listening in we have not we do not have a date for the tests net or the main net hard fork because we're taking we're taking very care to taking care of the fact that we need to be very mindful and careful and security conscious of the upgrades that we're doing good job soon thank you I am NOT a trained PR person

just throwing that out there does anyone else have any other comments that didn't make it on the agenda any other topics or things they want to say or discuss otherwise I think we're done

yeah while one

Péter Szilágyi: one thing I think we should do or would be nice to do is to I wanted to do this for Constantinople tube is to do a shadow for cover-up stone so essentially don't hard-code the fork block number on Robson rather just pick

one and run a few clients and a few minors with that block number that would shadow the real test nut and just see see what how things behave

Danno Ferrin: how would we keep that from corrupting the main Testament when we change the network ID on it

well

Péter Szilágyi: honestly I would just leave them as is so the two things would be connected and then they which have incompatible blocks to one another

right Rothstein's got enough issues okay as long as the shadow fork does not actually have higher difficulty or mining capacity then the main it rocks

them it shouldn't be too bothersome for the regular Robson right

Péter Szilágyi: yeah probably the problem yeah maybe there will be also a problem with synchronization because none of the one of the yeah keys actually change difficulty or headers so

for example you cannot fast think I guess the shadow drops them so yeah maybe but that could definitely right

Danno Ferrin: that's a problem we always see we only see all sorts of zombie chain forks coming from the Constantinople issues when we try and fast see so that's my concern I'm doing a shadow fork because we're creating more these zombie you know for

committed to take the notes down after while that's fine that some might sink to them

Hudson Jameson: are there any platforms that can simulate blockchain with traffic cuz I know white block has said before they have that and they open sourced it but I don't know of anyone who has used it outside of white block people I was thinking maybe consensus people might have but I'm not positive

Péter Szilágyi: well one other thing that we could try to do to do this

whole shadow work in a more controlled way is we could spin up a small cluster just a few machines running maybe one client from each team or each one of each client and instead of connecting them to Rob's done to have them completely firewood off of Robson and just have a transaction relay that just takes that connects to ops and actually is a full ops and oath and just pushes the transactions into this internal thing and then we can we can do whatever we want without actually being afraid of spamming blocks interrupts them back

Danno Ferrin: right we turn off your discovery and we only connected nodes with each other

that's a good solution I think

well I

Péter Szilágyi: would go a bit further and define some I would just start up a few cloud of VMs and somehow firewall them off of each other just to make sure that no client can be misbehave

yeah so

I'm pretty sure white block could do exactly this but I'll pass this along to Zach and I'm sure he'd be interested in helping in some way if we can

Hudson Jameson: cool so we have white block we have Peters idea I like Peters idea a lot because that seems very controlled in a way that we can quickly identify if something goes wrong without having to mess with like seeing if Rob's

tan how bad Rob stand breaks and having to do analysis on Rob's turn itself am I getting that right

Péter Szilágyi: mm so I'm not sure I follow the question so the whole shadow fork idea was specifically not to break Roxton prematurely but have have a cross client Tesla that's actually has traffic the bridge thingy that's more more or less the Protectorate stuff from this shadow network

Hudson Jameson: got it got it okay thanks that's where I was I was disconnecting there okay do we want to decide to just do that right now at minimum the shadow network idea

I'm

Péter Szilágyi: not

sure if we can speak as much there spinning up some virtual machines and deploying some clients on them that I mean I don't know how much time it takes but I assume it's not that big of a deal like somebody knows how to do it so maybe we could ask refers to spin up a tiny ansible cluster that's for the bridge itself it's actually what we need is erupts the node that that is a big act so that the transaction pool also forwards transactions to some external entity I mean if you guys want I can try to hack it on Monday and see if it's viable or not

Hudson Jameson: yeah that'd be really nice

James Hancock: I just got a new computer for my birthday so I could turn my old computer

into something do running one of those

taking

Hudson Jameson: it old-school we don't need no damn cloud of EMS all right yeah so we'll see what Peter does on Monday we'll get message back from white block and yeah happy belated birthday James I just got a message about that up James's unmute we're not gonna sing happy birthday because the lag time on zoom would make it very very bad and I'm

not a good singer yeah please please no okay anybody else have comments before we on the call oh when should the next meeting day right Oh Peter you go ahead first yeah so I don't think we can pick the block before the 6th am i right or is that six just when all the implementations will be done but we can have a block slightly before then

you

Péter Szilágyi: wouldn't pick a block number so I would say that if if parity

managers to implement or we have a fair consensus of implementations by next Friday then maybe we can talk on chat but as long as we don't have things implemented I mean we might as well decide today then

I see what you're saying so let's wait two weeks that does that sound good or were you actually were you propose I side are we saying that deciding today or in one week is the same if we don't have the implementation done so let's wait until the implementation is done great everyone good with two weeks from now that works good because we could stick

with the two week schedule we thread the needle of DEFCON 2

Hudson Jameson: okay and what time is the next meeting gonna be

that would be I think 22 UTC all right I think we did you like 14 then 22 and 6 no wait the one from last week was in the middle of the night listen you know yeah yeah yeah

Hudson Jameson: it was 22 last week and then this week was 14 and so we need to do this 6 6 we

need to discuss in the future changing this because we have so little people joining now from all the time shifting

yeah

Hudson Jameson: you know we actually could do you want to just have this at the same time in two weeks

because I think yeah

Tim Beiko: so can we say it's at the same time in two weeks and if anyone sort of feels strongly we should keep rotating because then they can kind of make a case for it

and we can start rotating again after that but personally like since we've started doing this it seems there's been like make minor benefits and major like detriment spikes I mean one time you know like you get thought the call was at one time and it was at another time and then they come in a couple hours after and it obviously makes for a less but not good call so it seems like that's been aside from the schedule just like having everyone's been so many mental resources on figuring out when the call is has been pretty overwhelming so I would

personally push to just have the regular time and if someone feels that it's really valuable to rotate and they can step up and you can say it either on this collar or elsewhere

yeah sounds great to me

Hudson Jameson: all right let's drop off this thing thanks everyone for coming and see you in two weeks at the same time 1,400 UTC

oh wait did someone have something yeah I just wanted to say that so I don't know

Péter Szilágyi: Martin suggested that we should ask whether we wanted to do a new test net

and essentially I didn't want to derail that discussion completely and say that we should do a foreshadow fork but I think people should maybe legitimately consider whether we want to reboot drops and maybe after a stumble because Robson is kind of big and and I don't think it's worthwhile to keep bringing that luggage along

much longer

Danno Ferrin: I agree with this one of the big issues were gonna come up with this product how launches as we're gonna use we're gonna need to do it a fork transition

so so well yeah but that's I mean

Péter Szilágyi: that should be fairly

simple so we can so since prop now is just empty blossom only needs empty blocks you don't need transaction so you can always pin up an empty Network and see

if the transition works anything else oh yeah we did cool anything else with the idea of you to restart Robson on the same t90 at zero or picking a chain ID for a second group of what training no I would can you run it I mean

Péter Szilágyi: if somebody wants to keep Robson alive I don't see a reason to explicitly murder it

just eventually the difficulty

balm will get it we don't forget yeah

Péter Szilágyi: yeah

Hudson Jameson: that gives people time to court to transition though there's some people who rely on Rob stand not that they should but they do

yeah of course that's what I'm saying that we don't

Péter Szilágyi: I mean I don't want to explicitly murder Roxton just if we want if we want to restart a new test that we can create something else and then eventually people can migrate and and we will just tell everybody that person won't receive the forks anymore so period transition when you can

sort of annual I mean it will be interesting the fraudster never got prop ow and the new network was one that got the product out that would be one differentiator

Hudson Jameson: oh yeah

totally okay any other comments all right we don't murder test nets we only make new ones

by everyone but I owe everyone great job Hudson thank you [Music] [Music]

[Music] [Music] [Music] [Music] [Applause]

[Music] [Music]

[Music]

Speaker names adapted from ethereum/pm, CC BY-SA 3.0.