Meeting minutes
<ottomorac> transcriber-bot, resume
Manu Sporny: He'd be open to that...
Otto Mora: Yeah, let me… Let me do...
Agenda Review, Introductions
Wilmer Daza: Sorry, guys, I'm in the car, so… yeah, my name is Wilmer, and I'm you in the group, and yeah, I'm the co-founder of My OnePlace, which is a company that is trying to create a...
… Universal layer, identity layer, and compliance layer for tokenized assets
… So, of course, very interested in anything which is identity for
… For on-chain, that was… Foundation
… I joined recently
Will Abramson: Great, welcome...
Stephen Curran: Welcome...
Wilmer Daza: Thank you. Yeah, you have any questions on...
Otto Mora: Okay, excellent. Yeah, welcome to the group, excited to have you on. Excellent. Uh...
Will Abramson: Yeah, I think I would just add, I mean, maybe I should, but, you know, if you have any questions for us, Wilma, like, you know, feel free to reach out to me or Otto. We would love to have someone who is...
… You know, another active participant, and if we can do anything to help you become more active and contributing
… Great
Wilmer Daza: Yeah, sure. I won't hesitate. I have a couple of questions already, so in the couple of weeks or days coming, I will be, like, really active. I've been seeing some...
… Because that I'm really, really interested in, and what are a great way, which is contributing to standard price, so that's better
Will Abramson: Fantastic...
Otto Mora: Fantastic...
… Oh, yeah, so, exactly, just the four topics, I guess, that we had there, but uh
… Um… yeah, do we wanna… Just go over those, I guess, Will, or… Yeah, go ahead
Will Abramson: Uh, do you have a side topic with maybe just, like, as a good intro thing, or, you know, some...
… we can just spend a minute or two on it. I… I wonder if people
… Have been using the minutes, and what they think of them
… I just say this because just before this call, I tried to read last call's minutes
… and I really didn't find them very useful. I mean, the problem is, they're very verbose, there's a lot there, and
… You know, we miss that human touch of, like, synthesizing the important bit, um
… so I don't know if there's a way that we can improve them, or… my sense was just
… Jesus, this is hard, this is hard to wade through, and there's some mistakes in there that, you know, like, names are wrong and stuff. So, I just wondered what people think
Otto Mora: I'll let my own, and then skip it after that, but...
Stephen Curran: All right...
Manu Sporny: Yeah, I mean, plus one to that. The way that we address that with the Google Meet infrastructure is, uh, I gave it very specific instructions to summarize...
… So, summarize all the topics that were discussed, provide a single sentence description of it, and then summarize the actions that were taken during the call, and make sure that you do it in, you know, a half a page
Will Abramson: Um...
Manu Sporny: Uh, that is the most… typically, that's what I've been using with those other… other things. I think that'll close the gap...
Will Abramson: Mm-hmm...
Manu Sporny: Um, it's very… I think it's… it is very important that we have the… the full transcription. I mean, yes, it would be much better if a human was doing it, and… and I did, um...
… uh, experiment with telling the LLM to, like
Otto Mora: Mm-hmm...
Will Abramson: Yeah...
Manu Sporny: Summarize, and sometimes it was amazing, and sometimes it was like, wow, that is definitely not what that person said, and that, you know, would get the person in trouble if...
Will Abramson: Mm-hmm...
… Hmm
Manu Sporny: you know, that was scribes. So, I think just keep it a transcription, uh, you know, level for now, um, and then do an auto summary at the top, and I think we're… we'd be good...
Otto Mora: Uh, Steven...
Stephen Curran: Um, I did a… a pass-it-through where I said, um...
… stick all of the statements from a single person in a… in sequence together, so removing all the ums and the ahs and things like that, it did a very good job
… Um, of doing that. So no summarizing, but simply removing the hesitations and things and combining
<TallTed> +1 adding auto-summaries will help. also cleaning out the "filler" non-words.
Stephen Curran: each entry into a single entry for a single person until it switches over to somebody else, and I found it did a really good job. I agree, a summary
… top and bottom, but I wouldn't have it in line summarize what anyone said
Otto Mora: Yeah...
<ottomorac> https://
Otto Mora: Let me see if anybody else wanted to… yeah, okay, so I'll acknowledge myself here
… Yes, um, we do have a list of open issues for the live archiver. One of the
… ones that I think, uh, right out of the gate, um
<ottomorac> ottomorac/
Otto Mora: Ted had pointed out, and I think a few others here as well, is this one around
… The hesitations and the ums and all, okay, and you know, and all these, like, I guess
… yeah, filler words, um, where… at least some post-processing. I've discussed this with, um
… Here, that does make sense to do at least just some basic cleanup for that, that would be good. And
… And then, yes, like, that
… That could be a good way to at least start some of that cleanup. Uh, yes, welcome to maintenance, Grant. You make yourself useful now, yeah. Absolutely, yes
… Uh, so I, yeah, I think both of those things would be good, uh, and then, yes, maybe
… keeping the verbatim, like, like you were saying, you folks were saying, I think makes sense as well
… But yeah, maybe then it's just a kind of exec summary type thing at the top that would make this more readable
… You know, with the potential for some… Minor, you know, like, uh
… What do you call it, uh, hallucination summary risk, but uh… that's just… yeah
… Uh, yes, Will
<TallTed> if the minutes go into git or similar management, instead of just keeping the "finished product", that will also be helpful -- for the inevitable accidental or malicious change history
Will Abramson: Yeah, I think that all sounds good. I think there's definitely ways to improve it. I mean, one of the things I'm wondering about is maybe there's, like, a post-post-post...
… processing script, you know, can be run is more deterministic than getting the AI to, kind of
… Edit out, but, like, there could be a script that removes the ums, for example, that is very
… You know, concrete, and obviously maybe we save the archive or whatever, but we don't render that for people to view
Otto Mora: Mm-hmm...
Will Abramson: Uh, and I do think it seems like the W3C is very interested in this tool, and hopefully they'll...
… you know, help… help you, Otto, or, like, help figure out an evolution of this tool that's better. And then also, maybe there's some way
… to, like, pass in, you know, like, we… the W3C has a lot of terminology, right? Like, um
… that is already defined in some way. There must be a way to feed that into the AI so that it
… Can, you know, draw on those terms, or, like, fit to those terms when it doesn't know, instead of, um… Just making stuff up
Otto Mora: wanted to jump in here to say that's what exactly I think would be the ideal summarization bot, if I may...
… Is a bot that's trained on the terminology and the context
… I don't know if it's just using the resolution spec and the DID spec as, like, the
… Be base vocabulary or something like that, but yeah, like, some, like, constrained bot that has a specific, uh, context
… Um, would be really good, yeah
TallTed // Ted (he/him) Thibodeau Jr (OpenLinkSw.com): Yeah, as I just threw into the chat, um...
… getting this into Git or similar management will also be a useful thing. Right now, we've got the IRC log
… Which is considered the rawest documentation. And then we have the parsed and pretty minutes doc that gets produced by
… The old bots, um
… And now there's somewhere in between those things what the… the transcription bot is doing
… Um, and what it's producing now as the rawest IRC log is not necessarily the rawest. content that it could be. Um
… As you were just talking about the, um… the W3, uh, language of its own, the cant of W3C, um
… The bots don't know that yet, but the humans participating in the meeting and taking minutes usually do
Otto Mora: Yes. Yeah...
TallTed // Ted (he/him) Thibodeau Jr (OpenLinkSw.com): And that's going to produce different output, right? So, there's sort of a middle ground that I think we're going to need to reach, where...
… The rawest possible output goes in as the base document, and then changes to it, whether they're made by bots or humans, go in as PRs that
… humans review and say, yes, that's a good change, no, that's crazy, what are you thinking?
… And that will lead us to a better place. I don't think removing humans from the loop completely is possible, and I don't think it's desirable. Uh, it's… it's the biggest
… Headache to remove is the actual
… taking notes during the meeting and turning that into its transcription bot is probably the biggest help we can do
… And then going from there to human reviews and human edits
… will get us to the next best output for the people, like, who can't meet the meeting, and those people read the minutes and get something out of it, while the people who are in the meeting
… Don't necessarily want to get out of it the same information, but they are in the best position to correct hallucinations or other issues based on the botch output. That's it
Otto Mora: 100%. One of my favorite bands, a British band called Public Service Broadcasting, so it quotes this old, like, BBC thing from, like, the 50s or 40s that said...
… machines will do the heavy work, and men will supervise the machines. And I feel like we're at that moment, like, the grunt work is being done by the AI
… Uh, for transcription, and now we need to do the supervision and cleaning up. So, yeah. Alright
… Yeah, I feel like that topic
… Yeah, let me just get with Pierre and see where we can do, at least for the cleanup of the homes and stuff like that. And then, yes, we do need to start to eventually train in AI that
Will Abramson: Yeah, I think we can move on now. I wanted to talk about the...
Agenda
Will Abramson: agenda a little bit, right? Otto, so we have… me and Otto put on the agenda 4 items
… But really, let's take a few moments now to talk about inner group, like, if there are things that you specifically want to talk about or dig into, we should do that
… Uh, the four items on the agenda are remove the dereferencing function signature
… So that's kind of where we were at last call. We were talking through that, and we made good progress, I think. We didn't get to a final discussion, so we could continue that discussion if there's things people
… don't understand about what's been discussed there, like, we're talking about maybe removing the function signature entirely
… And replacing it with an algorithm that maybe has inputs and outputs, that's another thing that's under contention
… Uh, and then the next one was, like, the HTTPS binding for dereferencing discussion, like
… If there are things we wanted to talk about that
… Uh, this other one, which we haven't really got to yet in our discussion, is passing a dig URL into the resolve function
… Uh, happy to talk about that, too. I think the… I think the set… my sense was from the group, most people are happy to do that, it's just Marcus who's very opposed to it, so maybe that's not the… best use of our time. And then the last one is
… Again, we haven't really had time to move on to it, but it's talking through the dereferencing algorithm
… My sense from the group is there is consensus that we should refactor this algorithm
… I think there is not concerns about what that refactoring might look like, and Stephen, I think that is… that is where the pass service
… Bit fits in for me, at least. Um
… Right, like, how does pass service get in this new algorithm? Where does it go?
<ottomorac> 1 - Remove dereferencing function signature
<ottomorac> 2 - HTTPS Binding for dereferencing
<ottomorac> 3 - Passing DIDURL into resolve function
<ottomorac> 4 - Dereferencing algorithm (PathService)
Will Abramson: So, yeah, I mean, I opened the queue. I'd really like to hear, is there… I mean, we could just go and spend, like, we have on the agenda 10 minutes for each
… Um, could do that, or
… Or we… if there's things or questions that people specifically want to talk about, or if there's other things that you think we really should spend some time on. Uh
Manu Sporny: Let's just… let's just jump into it...
Will Abramson: Okay...
Remove dereferencing function signature
Otto Mora: Okay. Pick one...
… Remove the referencing function signatures
… Yes, madam
Manu Sporny: Alright, so, uh, this is… this is the… this is not the HTTPS… Signature, right? Um...
Will Abramson: No...
Manu Sporny: Okay. Uh, I don't think this does… I think this is a big nothing burger, right? Like, why… I don't understand why people are… are...
… concerned about the function signature. Um, uh, if we just put it to an algorithm, you still have inputs and outputs
… Um, and we say these are the minimum inputs, and these are the minimum outputs, and we leave it at that. I don't… so… so I… I would like… and again, I would like to kind of understand why
… um, Marcus is pushing back so, so strongly on this. Um, and then, and then, you know, if we go to, like, oh, you know, someone's gonna object over it, it's like, okay, what exactly are you gonna object over?
… Like, you can't… you can't abuse Jack over some… if the… if the functionality is, you know, still there, so
Otto Mora: Mm-hmm. Well...
Will Abramson: Yeah, my understanding of one of the points of controversy is, um...
… I think Joe would say that there is no required outputs
… Alright, like, that maybe the way to say that in your language, Manu, is the minimum outputs from this algorithm is… is nothing. Right? Like, the algorithm can
… Do the dereferencing, and then use the dereferencing within that algorithm, without having to return it to some other. system. Uh
… But maybe that gets rounded by saying, like, the minimum output. And then I think to speak to, like, what I hear Mark is saying, it's that
… You know, if we remove the dereference function signature
… uh, I think he's worried that then that also rolls up into, like, HTTPS binding, and, like, there are
… People that depend on, you know, that's a point of interoperability. Option system. Uh, a smile to something
Otto Mora: Right...
Manu Sporny: Yeah, I mean, I would probably end up pushing back on Joe's position, because it's like having a function that doesn't return anything, and instead, it's like some weird state. I mean, you're basically...
… saving state in the function at that point. Like, if it doesn't return anything, then… It is… it is having a side effect
… Uh… that's… that's undefined. It's a useless function. Um
… So, I mean, maybe Joe's got, you know, something more specific there, but, like
… You know, you use the dere… you know
… It's a weird thing to define in a spec, um, to have a function that returns nothing. It has a side effect
Will Abramson: Hmm...
Otto Mora: Mm-hmm...
… Okay
Will Abramson: Yeah, I guess I can see that, but I think the example that Joe gave last time, which I don't have to hand, should do, apologies for that, but it was, you know...
… a very simple, like, dereferencing a DID URL with a fragment to a verification method
… Right, and dereferencing it in the context of, like, I want to verify a
… A verifiable presentation, or, like, approve on a credential
… Um, and his… the function that he gave, that, you know, like, the dereferencing operation was happening in that function, and then the verification of the credential was also happening in there
… And… I guess the return from that function that he, you know, like, mocked up
… Was, like, yes, the proof is true, you know, comes back as verified. I guess there is some argument
Manu Sporny: Yeah, but within… no, no, that doesn't… within the function, it had to return the document...
Will Abramson: Right, yeah, yeah, so it's where you draw the boundary on that. you know, I… Yeah...
Manu Sporny: Right, but I mean, that is the dereferencing function, like, that is the algorithm, so it does have an output...
Will Abramson: Yeah...
Manu Sporny: I, you know, if that's the case, I'd push back and be like, no, man, that's got an output, it's the DID document...
Will Abramson: Yeah. Yep...
Manu Sporny: Right? Or the verification method, or whatever, but it's definitely got an output...
Will Abramson: Yes, and, and, like, I think...
… We could say that, you know, that's still conformant. You've got an output, like, it's just not, like… there's not a clear, like, input and output to your, like, concrete function. That you've defined in code. dereferencing still
Manu Sporny: I'm trying to… I'm trying to find… I'm trying to find his exact example. I think it's in… it's in GitHub somewhere, right?...
Will Abramson: Uh, I think it's in the minutes from last week...
Otto Mora: minutes from last week...
Manu Sporny: There's our thing, and it's the past meetings...
Will Abramson: Oh, I've got it here...
<Wip> <JoeAndrieu> function checkProof(didUrl, proof) {
<Wip> <JoeAndrieu> let baseDid = removeFragement(didUrl);
<Wip> <JoeAndrieu> let fragment = getFragment(didUrl);
<Wip> <JoeAndrieu> let didDoc = resolve(baseDid);
<Wip> <JoeAndrieu> let key=getKey(didDoc,fragment);
<Wip> <JoeAndrieu> let result = checkProof_(proof, key);
<Wip> <JoeAndrieu> return result;
Manu Sporny: This was 23rd of April. Read minutes...
Will Abramson: I've got it here. I'll drop it in...
<Wip> <JoeAndrieu> }
Manu Sporny: Okay...
Will Abramson: Uh, and then just copy it...
Otto Mora: Oh, yeah, okay...
Will Abramson: So, I think it's really this get key. That is...
… It's kind of, like, the key is the
… Result of the referencing?
Manu Sporny: Yeah, but I mean, let's see, did Doc and fragment...
… I mean, he is dereferencing it, and he's getting a key back. That is the dereferencing function. GetKey is the dereferencing function
Will Abramson: Mm. Hmm...
Manu Sporny: Right? And he's getting a return value back, which is the key...
Will Abramson: Well, I would say that get key is one step of the dereferencing, right? Like, you've taken in this dig URL, then you've resolved the document. menu. I guess that, yeah, then you'd be referenced on the… On the returned bid document...
… Mm
Manu Sporny: Yeah, the other way to go about this is to basically say, you can implement this however you want to, and Joe, you've decided to implement it in a way that, you know, other people might not, um, and it's totally fine and legitimate, because you don't have to implement the exact algorithm that's in the specification...
… So we just put in language that says you can implement this in whatever way you want, as long as the end result is the same, which is
… you end up getting the key, and it's valid, and that sort of thing, right?
Otto Mora: I think...
Manu Sporny: and then… then we just get around it, you know, that way, I think. Joe's objection...
Otto Mora: I think Steven… Steven wants to jump in here...
Manu Sporny: Oops, sorry...
Stephen Curran: And just briefly, um… I mean, it all depends on how you define when does dereferencing end, so...
… You could say, yeah, you get the did document back, and then you get this fragment that points you to a verification method
… And there's a definition for what you do with, uh
… a verification method, because that's part of the spec as well, right? That's… that there. And so
… you know, Manu, what you said was the dereferencing stop when you pass the did doc back, and the client knows what to do
… Um, Joe's is
… You not only process the fragment, but you know what to do once you have processed it. If it points to a service, you process that service in the way that the
… Definition of that service is… is
… Specified. So, again, that's where, to me, it gets… it does get tricky
… Is, um, you know, where does dereferencing and. That's it
Otto Mora: Uh, yes, Will...
Will Abramson: Yeah, I mean, I have talked… heard Joe talk about that too, um, you know, like...
… You do reference and you use it within the context
… But… I think that's out of scope, almost
… for me, I think we can address Joe's concerns, and I think Joe would be happy, I don't want to put words in his mouth, like
… As long as that example that he gave, as one example of how you might implement dereferencing is not non-conformant, right?
… I think his concern reading the spec was, you know, it says I have to implement this function signature, and I don't want to do that, or I could imagine not wanting to do that. And I shouldn't have to do that
… Uh, the spectrum tell me how to do it. It shouldn't prescribe the way that I find the reference
Otto Mora: Right. Uh-oh...
Manu Sporny: Yeah, I mean, one way to approach this is, uh, when a group can't come to consensus on what the algorithm should do, you just don't write anything down...
… like, that's the end result. We didn't come to consensus, we're not gonna write anything down. And plus one will, you know, I think
… Addressing Joe's issue is just adding language saying, you can implement this algorithm however you'd like, as long as the end result is the same
… Right? And we can stay very loose
… about what that means, um, without, I don't think, creating any great risk in the
… interoperability of the ecosystem, right? Because most people are using this stuff to, like, dereference, uh
… a verification method, and then use it to verify a signature. Like, if you get that wrong, the signature's gonna fail
… And you're gonna… you are going to, as a developer, have to do something about that, like, align with the other developers, otherwise
… your software is not interoperable, so they're… they're market forces that'll get people to do the right thing here. Um
… Uh, and then for… and to Steven's point, you know, if… like, I'm very hesitant of saying, like, oh yeah, now we're gonna do, you know, process the service part. I know the path service thing is, like, involved in there. I'm wondering, Steven, if we can break that out into a separate algorithm?
… and just say, like, here's how you, you know, if your service is a path service
… Then here's the algorithm for processing pad service, and if it's some other service, then
… you know, up to you. It's implementation-specific, or whatever. So I'm wondering if we can just draw the line somewhere
… Um, where, you know, processing the service stuff is a separate algorithm. Um, I don't know if that would help
Stephen Curran: Um, just a matter of interest, that's exactly what I tried to do. And the way I did it was try to say. there's certain...
… Um, algorithms… there's certain extensions, or there's certain, um, definitions that are part of this spec, and you have to follow those
… And there's other ones that are extensions, and when you encounter those, look at the extension algorithm
… But yeah, I agree completely, that's what we've got to be able to get to, which is
<Zakim> manu, you wanted to ask if anyone is objecting to breaking out service processing into another algorithm?
Stephen Curran: You're gonna get something back, and then here's the… and then there's a definition somewhere, and it might be in this specification, it might be in the… did method, or it might be in the extensions, um, repository. That says what you should do with it
Otto Mora: Mm-hmm. I know...
Manu Sporny: Okay, so, I mean, plus one to that, Steven, so is anybody objecting to breaking it out?...
… I know that people are objecting to the details in it, but, like, what about… I mean, you know, going just from a top-level architecture perspective, like
… Can we have a dereferencing algorithm, and can we have a path, you know, service resolution? Algorithm separately
Will Abramson: Uh, yeah, I think at the high level, nobody is objecting to that, as far as I can understand, but I think the devil is in the detail...
… And my understanding of Joe's concern was really about
… How the referencing algorithm
… Defines, or selects, or identifies the particular service
… That it should be using to dereference a… it URL. I think it's the
Manu Sporny: I didn't quite follow that, Will...
Will Abramson: So, like...
… one… like, so I have a did URL, what does that URL target? Like, if it's targeting a past service
… like, how do I know that? And I think also, like, path services, like, path-based matching, right? Like, you match
… on the service, uh, based on the base path, or something like that, if I remember rightly
… And I think they're saying, well, there could be other ways that… or there could be other things that often would match against that path, potentially
… Um, and who has priority, and also, like, you can't
… there's something around there, right? Like, what is the ordering? Like
… I think Joe has said it would be best to first decide, like, first have some algorithm by which you
… sort of run through. There was some ordering, which you go, okay, first I'm going to check this, then I'm going to check that, and
… Uh… that's my understanding, at least. Um
… in the refactored approach that Joe provided, it was a sort of, like, step, you know, look in these places first, then look in these, then look in there
Manu Sporny: Yeah, I mean, I thought, Steven, you had already done that in your PR, um, and two, if Joe's done that, then that's fine. Like, I don't think… I don't think there's any pushback on that either. Like, yeah, there's… it's an algorithm, you have to do things in a certain order...
… And decide what you're gonna do, you know, as a pre-processing step, and then do the things
… As a result, like, I don't… I don't think that's controversial, is it?
Otto Mora: Uh, yes, correct? Steven? Go ahead...
Stephen Curran: Um, I don't think it's controversial that we do that. I think settling on what to do...
… Is still to be determined, but we haven't even got to that conversation, because we can't get there because of the. The holdup on
… you know, what is dereferencing? Um, so that's, uh… I think absolutely that's where we're going to
… You know, Joe disagrees with… I think… I think the positions are Joe disagrees with how I've laid it out
… And… and, um, I definitely don't think
… His algorithm is complete, or… or
… Or are necessarily accurate in certain places, so… but we haven't got to that conversation
Otto Mora: Yes, well… Oh, sorry, actually, my notes first...
Manu Sporny: Oh, right, so to get to that conversation, we just need to figure out where the line… lines are, right?...
Stephen Curran: Mm-hmm...
Manu Sporny: Like, and the lines… and they're just the interfaces, like, you know, where does the… what we're calling the dereferencing...
… You know, the high-level dereferencing algorithm start and end, and where does the sub-algram that does path processing
… You know, what are the inputs and outputs to that? And the previous, you know, algorithm calls the secondary algorithm, maybe, at some point
… And if we can't decide on what the secondary path referencing algorithm is, we're gonna end up specifying nothing in the spec and move on
Otto Mora: Well...
Stephen Curran: Mm-hmm...
Manu Sporny: I think that's the… right, right? I mean, how many different algorithms are we talking about here? At least two...
… like, are we just drawing the line on at least two? Like, there's the dereferencing algorithm, and then there's the path… You know, processing algorithm. Whatever we want to call it
Will Abramson: Yeah. Yeah...
Stephen Curran: Um… Right now...
Will Abramson: Uh, how you guys do?...
… There you go
Stephen Curran: Yeah, sorry, I jumped on. Um...
… Right now, there's, um, the… existing
… query parameters that were put in the spec long ago. Service, service type. And relative ref
… And there's PATH, which has essentially been
… ignored. I mean, it's there in the current spec, but it's… it's basically said, do whatever you want with it. Um, based on the DID method
… Um, and that's what we're trying… that's what the PAT service one tried to come to grips with, is let's define something for what
… path handling is, and let's do it… and I try to do it in a way that takes into account. Service and service type
… And relative rep, all 3 of those, in a way that's consistent with what it was done before, and
… And adds path handling to it. That's it
Otto Mora: Well, were you gonna say something, or should I let Manu go ahead...
Will Abramson: Um, the only thing I was gonna say is, I think… I think we just need to spend more time, like, as a group, pulling apart the… you know, like, breaking apart the algorithm as it is today, because. I think it is a bit...
… complicated, right? Like, that one flow. I know Joe's algorithm is not complete, but I do like
… The, sort of, like, breaking out into sub-steps that can be easily read, and then the third, this step does this
… And it… I liked that for readability, like, I think the current dereferencing algorithm is
… is complicated as it is, and was even more complicated, Steven, when you added your, you know, the path handling. I tried to, like
… uh, address all Joe's concerns about the different types of, uh, paths. Handling things that could exist
Otto Mora: Okay, Manu?...
Manu Sporny: Yeah, I think, you know, we just need… I am hesitant to go with what Joe's proposed, because he has not put as much time and effort into it as Steven has. Steven's approached...
… Sounds to me like it's almost completely complete. I think Joe's disagreeing with that, but like
… We need to start with one of them, and then see what parts people want to change, right?
… someone's got to propose something concrete. And I think we do have semi… we have definitely a concrete proposal from Steven, uh, we have an incomplete proposal from Joe
… if Joe wants to put more effort into it and make a, you know, fully fleshed out
… Complete, you know, algorithm, that's fine, but we just need to start with one of them and go from there
… Right? Um, and I think, you know, if we start with Stevens, we then go to Joe and say, like, okay, Joe, where are you drawing the line? Where would we tease this apart?
… and he's just gotta propose something concrete, and so does Marcus, right? Um
… I think… I think the problem with… the problem with this is, like, people are not
… we're saying, well, we need to make a decision on X, and it's like… that is going to take forever for us to have all these discussions. Does somebody propose something that has some reasonable
… you know, subroutine, you know, interfaces, and we'll go from there, and then… and we will refine that, right? People will find problems with it, and we'll move
… little subroutines between one another. Um… I'd much rather… and it's… and it would be really, you know, much easier if somebody just did a high level, like, here are the subroutines, XY, and Z, and then
… If someone's like, well, where do you do this? Or where do you do that? Then you should be able to point to one of the subroutines
… And that be… that be it, right? I think we have to go top-down here, um, meaning, like
… Talk about the top-level functions, the inputs and outputs of each one
… You know, and just go from there. And then we can figure out the details within, you know, within the tiny bits of the algorithm in each subsection
… And as far as PRs are concerned
… someone should just raise a PR for the top level. You know, subroutines?
… Like, no details, so people can't nitpick forever on the details, just high level, these are the sections, these are the subroutines, these are the inputs, these are the outputs
… Do we agree or not? And if anybody starts talking about details, it's like, nope, we are not there yet, just
Otto Mora: Hmm...
Manu Sporny: your detail goes into subroutine. Do you agree or not? Like, I think that's the level of the discussion we need to have, not… not like, I disagree with the way this, you know, sub-sub-sub-sub-sub function is discussed, like, no...
… Let's not do that anymore, because that's not really been getting us anywhere the past couple of months
Otto Mora: Well...
Will Abramson: I agree...
Otto Mora: Oh, you're, you're far away from the speaker, Bill...
Will Abramson: Sorry, yeah, I completely agree with Manu. I do think Joe would take exception by saying something like, well, I've already done that...
… like, I've proposed some subprime, if you guys haven't reviewed it, something like that. So I just want to, um… Bear that in mind, but I
… Well, I think… I think, you know, like
… Moving out of the details is great, absolutely. We want to know, like, what are the algorithms, what are the… what are, like, the sub-algorithms? What order do they go in?
… I… I would love to see, uh, Steven, you propose a, like, high-level piece of that, or even, like, critique Joe's, or, you know, like, take… look at Joe's, propose your own, like
… If we could both get a set of, like, what together you two think the algorithms are, that would be a great first start. Great start, uh, rather
Otto Mora: Mm-hmm...
Will Abramson: Because I think that is where the disagreement is at, or the places of disagreement are. Like, I think once we get to the ordering...
… Um… I think the details actually won't be… I hope they won't be too contentious
Otto Mora: Uh, let me see, I put myself on the queue. I wanted to say… one thing that I think has been harming...
… Other people from being able to contribute, or to, like, uh
… give an opinion on doing things one way or the other, is that sometimes I feel like
… we put together these… we also say, the people that have been putting together PRs, they put together the PR, and then it's just… it's like a… like a huge
… you know, a number of little text changes here and there, and, like, I feel like GitHub
… the diff isn't… the GitHub diff interface isn't, like, the best… Like, way to, like
… visualize it. I feel like, you know, you could… I don't know if it's a side thing before you do the PR, or you do the PR, and then you do this other thing where it's
… You know, like, we had the Google Doc, and I felt like that was, like, really useful in people seeing, okay, issue number 1, 2, 3, 4, 5, 6, 10, right? All of that
… was, like, very, very visible. And if here, if it's, like, maybe some kind of similar kind of Google Doc, where this high-level function is proposed
… And then, you know, like, Joe's gonna nitpick something about it, or someone else, and then… Instead of, like
… the really long thread GitHub discussion gets really hard to follow. Like, just having the comment that highlights the specific line of the function that you're disagreeing with, or pointing, nitpicking something about
… Make… may make this a little more easier to have that discussion, because then when you're seeing that, and you've been screen sharing, that it might also
… give… lead to a more productive, like, more efficient conversation, but just my two cents on
<TallTed> I think better to use github wiki or .md in the relevant repo, than a detached google doc
Otto Mora: You know, one way to perhaps do it, so… yes
Manu Sporny: Um, I mean this in the most loving way as I can. Uh, the PRs are being done in the exact wrong way...
… they are… they… they… it is inevitable that they're ending up where they are, because they're massive PR attempts
Otto Mora: Mm-hmm...
Manu Sporny: please stop. Like, I have said this, I don't know how many times over the past 3 months, it is inevitable that you will not get a PR in if you change absolutely everything all at the same time. The way to do this, and I'm… and again, like...
… with all the love in the world for the people that are raising PRs, the way to do this
… is you do a very high-level PR that just talks about a function, you delete absolutely every step out of that function, and you say, these are the inputs, and these are the outputs for this function, and you raise a single PR for that thing
… That is the way that you make progress on contentious things. I feel, you know, I know I've been… I probably sound like I'm repeating myself for the 20th time, but
… that it is… the PRs are being done in the exact wrong way. Please stop doing these PRs in that way. And I'm sorry, Steven, it goes to you, it goes to Joe, it goes to Marcus
… These, like, endless threads over hundreds of lines of changes is just
… it's not… I mean, it's very clear. It's not getting us anywhere, and… and
… And it… and it… and this is… this is… this is so predictable, like, over
… You know, talking about, like, 18 years of editorial experience, this is inevitable when you approach PRs in this way, this is the outcome every single time. This is why I said months ago. Please don't do massive PRs. Um
… So, I would like to see, if we do raise PRs
… The PRs are limited to two paragraphs. The first PR is, here is the X function
… and that is a separate PR from here is the Y function
… And when you do, here is the X function, you only talk about the intro paragraph
… here are the inputs, and here are the outputs. And discussion is limited to that text. That's it. Like, don't bring in anything else into the discussion. Don't talk about the details of step one or step five or whatever
… That comes in a second PR, after we've already agreed to the interfaces
<TallTed> +1 manu -- and those high-level PRs should *leave intact* the stuff they mean to replace. Create as parallel sections to what already exist, not as replacement.
Manu Sporny: Sorry that I'm repeating myself for the 50th time
Otto Mora: No, no, no, all good, all good. Just in defense of Steven, I...
… I think he sort of came to that realization over time, right? Like, you did the PowerPoint eventually, right? But go ahead, Steve
<manu> sorry!
Stephen Curran: Hang on! So, I'm on the queue, and Natty, you just jumped the queue...
Otto Mora: Sorry. Sorry about that...
Stephen Curran: talk there. So I'm calling you out on that. Second, um, I was gonna repeat what Otto said, which is, I learned my lesson, I did one PR, saw it was too large, my current PR has two chunks of text in it, that's it...
… Yes, one of the chunks is long, but it is
… inserted, and that's it. It doesn't change anything else. Um, so I wish we could have talked to it, but we've really never talked to it
… Um… yeah, I agree. So, anyway, seriously, what I was gonna talk about was, um
… That… I have not really paid that much attention to Joe's
… Algorithm, because other than looking at it and saying, oh, that's not… Compatible with what we have, and not
… complete. Um, I've waited for us to get to the point where we're talking algorithm
… So, I'm happy to do that. I'll try to do exactly what Will suggested, which is
… Um, I think I did… what I tried to do was say, oh, here's what the
… algorithm was before, I'm… I want to stick this in it, here's how it can flow in, including that. I will try to do that with Joe's, and see if I can come up with something. Um, time is a little tight for me these days, but I will do my best to try to get that pass done. where I insert that
… But I think that is the right path. I just didn't know whether we were to the point where we could talk about the algorithm because of all the other issues around it, and I totally agree with the
… um, change everything, including all formatting, in a PR, is a real problem
… Uh, that's it
Otto Mora: Go ahead...
Manu Sporny: Yes, plus one to all that that Steven said, and I just put myself on the queue to apologize to Steven for jumping queue...
Otto Mora: I did too, because I wanted to defend...
Stephen Curran: Apologies accepted...
… All good
Otto Mora: I did too, because I wanted to defend… you did the research, you did do the PR, so then you started to break it down in smaller chunks, and then you even did the PowerPoint presentation, which I thought was really useful...
Stephen Curran: And I think that's important, is, um...
… It's really hard to come into a… now I've jumped ahead. No, go ahead. Manny, go
Otto Mora: It's...
Will Abramson: It's me...
Stephen Curran: I...
Will Abramson: I think it's me, but you can go, Steven. Please listen...
Otto Mora: We'll… we'll...
Stephen Curran: Yeah, okay, so one last point, which is, I do think that having… when a PR is reasonably large, requiring the author to put in a...
… you know, do a PowerPoint, present it, discuss it, why, and give people the contact before they go and try to pick it apart is a really good process. I wish we could do it more when it's large. Obviously, for little ones that are focused
<ottomorac> +1 to swcurran's suggestion
Stephen Curran: less important, but when, you know, everything changes, I think that's crucial
… That's it. Thank you. Rant over
Otto Mora: Well...
Will Abramson: Yeah, I would agree with that, Steven. I mean, I think it almost, like, before doing a big PR like that, there should be work to...
… get consensus on the big PR change, right? Like, which maybe hasn't really happened. I just wanted to speak to, like, um
… Joe's five points, right? Like, I've got his PR up now, and I'm sure there are bits that we disagree with, but, like, this is the kind of thing that I would like to see from
… you too, Steven, so he has, like, prepare to resolve the URL, resolve the dig URL. I think we can all agree those two steps
… you know, maybe we name them differently, but they definitely happen first, and I think there's consensus on that, or maybe people would say differently. And then he has determined retrieval strategy
… retrieve the resource, use the resource. So I think it's those last three, um
… And I would say, I think Joe's sort of been a bit blinkered focused on, like, handling paths
… in this case. So, you know, and maybe there's dereferencing that isn't
… I don't know, I just wanted to, you know, get those on to the point, and
… Um… I think, literally, if we could have just 5 points from you, Steven, like, okay, these are the names I would give these things, or… Uh, and I think, really, that the bit that we
… would then have to drill down in deeper. Is this, like, what do we mean by determined retrieval strategy? I think retrieve the results also is… Is, uh, fine. Anyway
Otto Mora: Mm-hmm. Uh, Steven...
Stephen Curran: Yeah, I mean, my reaction when I saw it was… I like the idea, but the...
… But things like breaking apart service and service type, like, and ignoring the fact that there could be other things
… that are kind of… that are extensions that are, like, where do you include… those are the places where I thought it was weak, and… and
… But again, I haven't really spent a lot of time on it because the conversation hasn't gotten there. Um, but I'm happy to look at it now
… Or try to
Otto Mora: Yes, ma'am...
Will Abramson: Yeah, great. Okay, well, maybe I'd suggest, like, we should… because I agree, we've been meaning to get to this, and I'm sorry, Steven, you did a lot of work on the past service, and then we kind of park this, like, this whole dereferencing, sort of...
… throw everything in the air and see what we think about it. But maybe, like, next Thursday, we should plan to just… we could have the whole
… call on the dereferencing algorithm, right? Maybe Wednesday we can
… Continue to try and, like, close out the final discussions, but
… I think we can still talk about the meat of that algorithm. I always start to have that conversation
… Um, because it's probably going to take a couple of calls. I think it would be great to get to it. because I think that is a place where… Um
… I think we can get to consensus on the structure of this algorithm. Um, I think there are less, sort of
… definitely opposing views here. You know, like, for example, the did URL being passed into the Resolve algorithm
… Marcus is very strongly opposed, right? So, like, getting through that is going to be a bit more, like, a conflicting one, whereas this, I think, can be a constructive conversation that will improve the spec. And it would be great to
<ottomorac> transcriber-bot, pause
Will Abramson: Have some positive conversation in the group. Board momentum
<ottomorac> transcriber-bot, resume
Otto Mora: Uh, mano… yes, man...
Manu Sporny: Yeah, I mean, I feel like a PowerPoint is a step backwards, like, we just need to get the text there. I'm not...
… fully opposed to it, but, like, maybe PowerPoint with the set of PRs
… Um, or we spend the time saying, like, okay, what are the inputs, what are the outputs? That's what we're talking about today. We're not talking about any of the details
… what are the various ways that we can create these interfaces? Like, what are the interfaces?
<swcurran> agree -- the powerpoint is not the end goal. A PR is needed, a powerpoint is to present the idea.
Manu Sporny: and just get… have a discussion around that. I mean, it's fine, I guess, if, you know, Steven, you propose something with a PowerPoint, we have a discussion about it. But it should very rapidly change into
… We have 3 interfaces, uh, we have 3 PRs
… And the PRs are a single paragraph introduction to what the function does, its inputs and its outputs. And that's it. Zero details. Like, we'll get to the details later
Otto Mora: Mm-hmm...
… Avoiding that temptation to have the grand vision and go into all these specifics, right?
Will Abramson: Yeah, I definitely agree, but I do think, like, a PowerPoint, like, something that frames this as, like, a collaborative discussion, where we, as a group, gonna get to consensus on those inputs, you know, like, what high level, we're not going to get into the details, but what are the steps, and, like, the algorithms and the inputs
and outputs?...
… But we can frame that around, and these are going to be, you know, we're just going to scaffold this out as PRs, like, the result of this conversation is going to get turned into PRs straight away. I'm definitely
… on board with that. My hesitation of, like, getting it straight in as a PR is… I think people who disagree will
… You know, come in a confrontational manner, as opposed to a construction
Otto Mora: Hey, not seeing any reactions… Um...
… Yes, I will plus one that going straight into PR might… I don't know if it's this, like, oh, this is gonna make it into the spec, I have to react, like, uh, just a knee-jerk reaction, but uh… Maybe… yeah. I don't know
Manu Sporny: Yeah, I think… Will, you make a fair point, so let's try that next time. I just want to make sure, by the end of the call, we are ready to write, you know, one PR for each. Interface that we discuss...
Will Abramson: Mm-hmm. Yep...
… Okay, so we should plan that for Thursday
Otto Mora: Okay, so Thursday, so no special topic call?...
Will Abramson: Um...
Otto Mora: Or, like, just to give time for people to, like, prepare...
Manu Sporny: We should have… we need to meet much more regularly than we're meeting to make progress on this, otherwise we're gonna run out of our 6-month extension before we know it...
… Like, we need to be done in a month and month and a half with this stuff
Will Abramson: Yeah. So, as I was thinking we'd meet Wednesday and just continue from where we left off last week, you know, trying to get just final clarity on, like, the high-level algorithm and, like, the inputs and outputs...
Otto Mora: Okay...
Will Abramson: And hopefully, uh, you know, like, HTTPS binding, we still need to, like, close that out...
… Um, I think Alex is compromised with the… and then also the URL. There are a few sort of small, but more contentious things that
… That we should just put in that Wednesday call, and they're not going to get talked about in the Thursday call, and then Thursday call, we can just have it focus on
… the, like, internals of the, uh, URLD referencing of
Otto Mora: Yep. Good. Perfect...
Will Abramson: Okay...
Otto Mora: Excellent. Thank you all. Appreciate everybody's feedback here. Thank you...
Will Abramson: Yeah, that's it...
<ottomorac> transcriber-bot, pause