Agoric SES, 10 Apr 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SES Meeting: Import Reflection

Description

Import Reflection proposal has import source syntax to obtain compiled (uninstantiated) module source, such as a WebAssembly.Module. Discussion of its interaction with CSP, Module layering (née Compartment proposal), and SES shim.

A

Michael fig:, good morning, everyone.. This is this meeting for march 20, ninth.

A

Michael fig: today we're going to cover some topics on javascript and port reflection, and also as a summary of the tc. 39 plenary. That happened last week.

A

Michael fig:, with that left hand, it off to guy has some ideas on the portal reflection.

B

Guy bedford: hi!, so we initiated this discussion itself out of tc. 39.. We presented the.

B

Guy bedford: import reflection proposal again at the last meeting, and based on a.

B

Guy bedford: the new simplification of the proposal as as representing phases of the learning cycle,, and we generally got a positive feedback on the specification, and it's something that we're looking to move towards progression. On.. So.

B

Guy bedford: we're. we're at the stage of getting stage, 3 of us for the spec., and we would like to be able to propose stage 3 at some point in in the coming months..

B

Guy bedford: and in order to in order to that, one of the things that was brought up at tc 39 was the sort of.

B

Guy bedford: question about how these module specifications into relate, and how they interrelate, with all these difference,.

B

Guy bedford:, virtualization, workflow. and so there's a lot of cross-cutting concerns that we're working through in the modules group meetings.. But it's nice to also discuss those with a compartment specific.

B

Guy bedford: focus,, and so I was happy to try and engage in more of those discussions here.

B

Guy bedford: and the at its at its core., the.

B

Guy bedford: import reflection, proposal is a new syntax for importing a module at the source face,, which means that it is compiled that it has not been linked.. The dependencies have not been fetched.

B

Guy bedford:- and it is a a monroe in the case of bloom that is compiled and can be used to instantiate, multiple times.

B

And we have very much the same.

B

Guy bedford: description of modules for javascript guy bedford:. In the discussions that we've been having around mo javascript virtualization,, which is that you have module source and module, representing.

B

Guy bedford: those same 2 phases, and so our expectation at the moment.

B

Guy bedford:, the the expectation would be that the corresponding import reflection for javascript would be a it would be import, and then the source keyword.

B

Guy bedford: and then the identifier that you want to use for the source. and if you're importing from webassembly, you would get the webassembly monitor. Object.

B

Guy bedford:, but if you're importing from javascript,, we should be able to get, give back the the module source object that we're expecting to be able to provide for javascript,, which again would be that compiled, but not linked or evaluated., and it can be still evaluated. Multiple times.

B

Guy bedford:, so guy bedford:, if, if we're potentially going for stage 3., what we're gonna do for now is is basically through for javascript source imports, because we don't have them yet.

B

Guy bedford:, but it would be nice to work towards a place where we know how those are going to come. Through. so.

B

Guy bedford:, we have an api for module source within the compartments proposals at the moment.. The question is just,. You know.

B

Guy bedford: is that api, something that, in its current form, we can expect we'll be able to support import reflection.

B

Guy bedford: and when we could expect it. and- and you know what what this,, what the process might,, how how it might turn out on that.. I think there's some really interesting virtualizations, that that can allow.

B

Guy bedford: and then that it sort of starts to reach into a whole bunch of the layering, questions, and and the conversation can certainly open up from there.. But that's that's. The overview. nicola can certainly provide more context, and I'm sure karate will have.

B

Guy bedford: a picture of how this fits,, together with compartments as well.

B

Guy bedford:, if either,, if you want to sure any thoughts, that'd be great.

C

Nicolò ribaudo : yeah,, just on that,, we like doing the mob to modules. Meetings. we've also discussed how to split.. Maybe the compartments proposal differently like to propose some other layering.

C

Nicolò ribaudo : after that nicolò ribaudo :. We can find some shared minimal piece. That would also be useful for the other proposals.

C

Nicolò ribaudo : because, for example,. It would make sense already. she. for example, module services for javascript,, even without all the input hooks.

C

Nicolò ribaudo :, you know., I remember hook already with.

C

Nicolò ribaudo :, the imperfection, proposal, or immediately after. so like our discussion,, is also like not only how it wasn't fit together,, but also what is the.

C

Nicolò ribaudo : like?, the best.

C

Nicolò ribaudo : the way to group them,, because we are both trying to avoid hard dependencies between proposals and making sure that. like,. Even if at any time one proposal gets locked like the.

C

Nicolò ribaudo :, the remaining parts are still consistent and usable.

D

Mark s. miller, (mm, agoric): yeah,, so I do. you have your slides, easily available.

B

Guy bedford: yeah,, let me get the link. I'll show that. now.

B

D

Mark s. miller, (mm,, agoric): yeah, and you know, we're on zoom.. So so you can screen, screen,, share the slides and and talk to them a little bit.. The the the.

D

Mark s. miller, (mm, agoric):, the 5 part pipeline of.

D

Mark s. miller, (mm, agoric): module processing, and how those relate to.

D

Mark s. miller, (mm, agoric):, the mark s. miller, (mm, agoric):, the proposed syntax,, I thought, was really very, very elegant.

D

Mark s. miller, (mm, agoric): reconceiving of of how they of okay. okay. the larger.

D

Mark s. miller, (mm, agoric):, you know, design space again for reflection, and how it all can fit together in a very.

D

Mark s. miller, (mm, agoric):, smooth matter.

B

Guy bedford: yeah. so put together., and this this was nicholas, actual reframing in terms of phases that this,, when we use this syntax.

B

Guy bedford: for the for the import.. What what we're using for this first keyword.

B

Guy bedford:, it represents a a phase of the pipeline. and- and so in this case, we're we're obtaining the module at the source phase, which is this fetch and compile phase of the instantiation pipeline.

B

Guy bedford: and effectively for webassembly that helps us avoid.

B

Guy bedford:, this kind of monstrosity, and I think, there's actually a webassembly that instantiate streaming that's supposed to be in here as well.

B

Guy bedford: that that.

B

Guy bedford:, currently you have to do today to load webassembly, and so the by having something that represents that phase directly. We turn what webassembly from something that's kind of very, you know, unwieldy to kind of analyze or for build tools to understand., and it's it's.

B

Guy bedford: sort of something- that's that's very much, not scared, first,. Last support to something that's very much, has a much.

B

Guy bedford: closer integration with the module system., and so this would work for less than without proposal. and then,. If you're, if you're learning a. js file, we're hoping to be able to achieve the same thing,, and then we potentially would have other phases,, possibly an asset phase, and.

B

Guy bedford: and and the the the execution and and linking phase.

B

Guy bedford: is represents the deferred imports proposal,, which is another proposal, that's happening at tc. 39,, which I'm happy to discuss as well. If there's interested in in in discussing that. there's, certainly some cross-counting relations as well,, but.

B

Guy bedford:, maybe on a slightly longer top line for them. karity. sorry! did you want to say so?.

E

Caridy: no, no,, I I was. I was.. I was going to ask for the one that you have really for or some other example beyond the the the source on me., so yeah, and pretty cool.

B

Guy bedford: yeah, it., it's been really nice that, instead of you know the the the risk with all having all of this stuff together is that it over complicates.. But it's nice when we do find.

B

Guy bedford: the simplification patents that that, treating everything as this epic that has had that benefit of being able to find the the the simpler patterns. and- and hopefully we can do that for.

B

Guy bedford: reflection as well,, and and have some good shared, simple patterns that can create some,. Some really strong.

B

Guy bedford: and powerful virtualization primitives that can make virtualization easy and and powerful in javascript., and I think this.

B

Guy bedford: some really exciting stuff that we can do there.

E

Caridy: that you were showing before.

E

caridy:, so this one,, not the one, that's the one that changed. so.

E

Caridy:, the in this particular case., the.

F

Caridy: they want the volume that you're gonna get there. It's going to be.

E

Caridy: the the thing that you call caridy: compile screen streaming with a compile with it?, or is it going to be already compiled.

E

Guy bedford:, by the time you get it? yeah.. So that's what the correction? sorry! I didn't mean to put instantiate streaming, that the correction to the slide is compile streaming.. So it's you're, getting the compiled source.

B

Guy bedford: so you, and you compile it once you can cache the compilation, and we use that compilation for all future instances of that module.

E

Caridy: okay., so so it you know that, because you're getting it was a module you're, gonna, get. and you're asking for the source. you're gonna get the compiled version of that.. So the value of what's awesome will be that case.

E

Caridy: a. a a compiled webassembly module.

B

Caridy: okay. guy bedford:, it's a it's a webassembly dot, module.

B

guy bedford:. There is some confusion in terms of a wording here, because.

B

Guy bedford:, you can think of this, as in in our terminology.. It's webassembly dot source module,, but it's called webassembly dot module today on the web., so that there is a confusion there. but.

B

Guy bedford: yes, webassembly dot module as a source.

C

Nicolò ribaudo :, like even with javascript affection, like in the model for the subject., I believe, browsers, could already have, for example, the byte code.

C

Nicolò ribaudo : on the monitor, subject without like,, instead of just being an object, container string.

D

Mark s. miller, (mm, agoric): yeah, nick.

D

mark s. miller, (mm, agoric):, one of the mark s. miller, (mm, agoric): well.. First of all,, I mean, you know, endorse the idea that we should just avoid the the unqualified.

D

Mark s. miller, (mm, agoric): work, module as as much as possible, and always., and you always use adjective module.

D

Mark s. miller, (mm, agoric):, the source module is, is what, yeah,. What we've been using for the result of.

D

Mark s. miller, (mm, agoric): of compiling a module source, and and it's.

D

Mark s. miller, (mm, agoric):, the it's., it's.

D

Mark s. miller, (mm, agoric): well named in one regard and misleadingly named in another regard.. It's well named in that.

D

Mark s. miller, (mm, agoric): it mark s. miller, (mm, agoric): only contains information derived from the source. there's no.

D

Mark s. miller, (mm, agoric): context, information inside the the module inside the source, module.

D

Mark s. miller, (mm, agoric):, it's misleading in the sense that the source is not recoverable from the source. Module.

D

Mark s. miller, (mm, agoric):, the so you know, and in in particular, this was very relevant to us for.

D

Mark s. miller, (mm, agoric):, what's going on with with mark s. miller, (mm, agoric): with multiple and excess.

D

mark s. miller, (mm, agoric):, where they mark s. miller, (mm, agoric): pre-compile their modules during development, and then.

D

Mark s. miller, (mm, agoric):, all that exists in the rom is the compiled modules.. The source information is purposely dropped.

D

Mark s. miller, (mm, agoric): now, of course, in in wisdom.

D

mark s. miller, (mm, agoric): a source doesn't mean source text in any case.

D

Mark s. miller, (mm, agoric):, but mark s. miller, (mm, agoric):. Let me first of all, just bring up immediately sort of what a concern that I've got. That does relate to one of the constraints that we've that, in our understanding, what a source module is.

D

Mark s. miller, (mm, agoric):, which is in this import reflection, proposal in the full.. You know the full proposal. the.

D

Mark s. miller, (mm, agoric):: this is.

D

Mark s. miller, (mm, agoric):, the import reflection is expressed with syntax. and.

D

Mark s. miller, (mm, agoric):, therefore.

D

Mark s. miller, (mm, agoric): it's.

D

Mark s. miller, (mm, agoric):, it's available to any you know.

D

Mark s. miller, (mm, agoric): to any. get any code. That's not.

D

Mark s. miller, (mm, agoric): there that we're not trying to filter out by parsing and the source module.

D

Mark s. miller, (mm, agoric):, fortunately for both javascript and for.

D

Mark s. miller, (mm, agoric):, is completely powerless.. It's a immutable,, powerless thing that really does just provide information. and and the you know, potential to turn into execution..

D

Mark s. miller, (mm, agoric):, but does not itself.

D

Mark s. miller, (mm, agoric): provide any ability to cause effects. The concern is that.

D

Mark s. miller, (mm, agoric):, whether the wisdom module is powerless.

D

Mark s. miller, (mm, agoric):, is not under the control of tc. 39. It's a separate standard.. Some.

D

Mark s. miller, (mm, agoric):, first of all, worried about. how do we. in in proposing this, and then integrating it into the ecosystem?.

D

Mark s. miller, (mm, agoric): what do we,, what you know?? How do we say? what constraints we're imposing, and to what degree are we just delegating to the wisdom, spec.

D

Mark s. miller, (mm,, agoric): and.

D

Mark s. miller, (mm,, agoric): and I'll start there.

B

Caridy: you've had your hand up for a while.. Is that on this topic, or do you,, should we discuss this topic first?? I I I wasn't in the meeting last week. I was on vacation,, but.

E

Caridy: and it. caridy: it.. It feels to me that I have a gap. On.

E

Caridy:, how are you going to use that module that you are importing there? and I- and it's not very clear to me specifically whether or not what mark is saying is.

E

Caridy:, actually, what caridy:, what this is doing., that they, the source doesn't, have any other information about the place where the source was actually.

E

Caridy: and so on.

B

Guy bedford: yeah.

C

Nicolò ribaudo :, so first the question for mark, to better understand this question is like.

C

Nicolò ribaudo :. How does this concern regarding imports.

C

Nicolò ribaudo :, using like nicolò ribaudo : arbitrary host objects such as we assembled module?.

C

Nicolò ribaudo : does it also apply to standard imports where hosts could have their own.

C

Nicolò ribaudo : built in models, and give you any object, or is it different and doing to protect.

C

Nicolò ribaudo :, in this case a different way?.

D

Mark s. miller, (mm, agoric):, so that that's a great question. is certainly the case that.

D

Mark s. miller, (mm, agoric):, regular imports are resolved by the host, and the host.

D

Mark s. miller, (mm, agoric): is. mark s. miller, (mm, agoric):, you know, free., you know the the free, according to the terms of the current spec., to provide.

D

Mark s. miller, (mm, agoric): a. you know, module instance that provides arbitrary I/o.

D

Mark s. miller, (mm, agoric):, the.

D

Mark s. miller, (mm, agoric):, so I,.

D

Mark s. miller, (mm,, agoric): and systems like the session emphasis on shim.

D

Mark s. miller, (mm, agoric):, which are trying to produce the security properties.

D

Mark s. miller, (mm, agoric): by,, you know, using curtis 8 magic lines of code trick.

D

Mark s. miller, (mm, agoric): and without.

D

Mark s. miller, (mm, agoric): do have to for for valuable strings.. We still have to scan them for things like.

D

Mark s. miller, (mm, agoric):, dynamic, import.

D

mark s. miller, (mm, agoric):, the you know,, the dynamic import expression, since they can occur in a valuable springs. and for modules. We do have to actually.

D

Mark s. miller, (mm, agoric): partisan, translate, to turn modules into a valuable strings in order to feed them into the magic lines of code..

D

Mark s. miller, (mm, agoric):, so I suppose. putting that whole picture. Together.

D

Mark s. miller, (mm, agoric): having this particular.

D

Mark s. miller, (mm, agoric): angle provide access to host, provided power, should that occur.

D

Mark s. miller, (mm, agoric):, does not present a danger of a different kind, and kind of already falls into the protections that we have to have.

D

Mark s. miller, (mm, agoric): for import in general.

D

Mark s. miller, (mm, agoric):, the.

D

Mark s. miller, (mm, agoric):, I think I'm convinced by that, so I'll I'll. I'll retract, that as a gating concern. in general,, it's we're we're trying to.

D

Mark s. miller, (mm, agoric): prevent access to new powers coming in from syntax., but in this case, I suppose the the argument.

D

Mark s. miller, (mm,, agoric): they're, we're talking ourselves into, is that it's.

D

Mark s. miller, (mm, agoric):, providing access to power mark s. miller, (mm, agoric): through a potential access to power, through an extension of syntax that already provides access to power. and all the mechanisms for protecting against that would also apply to this extension.

E

Caridy: okay, cool. caridy:, all right.. I I agree with that.

C

Nicolò ribaudo : okay, by the way like when discussing what's been for 19 per syntax for this someone he may be, chris, many months, ago.

C

Nicolò ribaudo : propose some like important source.

C

nicolò ribaudo :, and then we moved back to this, like just.

C

Nicolò ribaudo : put in the face nicolò ribaudo : in the options bank,, so that the existing regular expression that the ss. in uses like.

C

Nicolò ribaudo :, not the broadcast., but you dynamic in.

C

Nicolò ribaudo :, so like we. nicolò, ribaudo :, we are tried, like we have been using that for the static version. but, like we., when adopt not changing this index. For for this reason,, I believe.

C

Nicolò ribaudo :, but it was nicolò ribaudo : many months ago,. So it might be me some numbering.

D

Mark s. miller, (mm, agoric): okay,, so that that does sound like a good argument.. I think I do recall that argument being made yes.

D

Mark s. miller, (mm, agoric): to to argue on the other side the if we wanted to be able to statically distinguish.

D

Mark s. miller, (mm, agoric): the what would be the 5 different flavors of dynamic import?.

D

Mark s. miller, (mm, agoric):, the pseudo property on the import keyword.

D

Mark s. miller, (mm, agoric):, you know,, because import keyword is a keyword. and the the and the the properties or signal properties,. It would be a a hard static, syntactic difference.

D

Mark s. miller, (mm, agoric): that would reliably mark s. miller, (mm, agoric): distinguish., mark s. miller, (mm, agoric):, whereas if they're all folded together into just the dynamic import expression based on dynamic parameters., then you can't.

D

Mark s. miller, (mm, agoric): reliably, statically distinguish them.

G

Mathieu hofman: yeah,, except that it's a lot. it it's basically really really hard to.

G

Mathieu hofman: easily mathieu hofman: figure, out. import dot source for similar things. and I think we're source another problem. and I think it was more.

G

Mathieu hofman:, I I can't remember exactly,, but you can do it with a with a with a red x.. Basically, you have to go in like and and parse the code.

D

Mark s. miller, (mm, agoric):, I don't believe that.

D

Mark s. miller, (mm, agoric):, the I mean the red x that we're currently using for.

D

Mark s. miller, (mm, agoric):, the dynamic import expression., the price of a red x is.

D

Mark s. miller, (mm, agoric): that it will. mark s. miller, (mm, agoric):, you know, make errors of in,. You know, make errors of inclusion, the errors of of recognizing things that are not genuine import expressions, because they occur in comments, or in strings, or.

D

Mark s. miller, (mm, agoric): or whatever.

D

Mark s. miller, (mm, agoric):, the.

D

Mark s. miller, (mm, agoric):, the yeah, we, the import keyword,, followed by.

D

Mark s. miller, (mm, agoric):, you know, possible white space,, followed by a dot,, followed by possible white space, follow up on identifier.

D

Mark s. miller, (mm, agoric):, so you have., we have to. conservatively.. We have to make sure that we never miss one.

D

Mark s. miller, (mm, agoric):. So what do we do for the? I think.

D

Mark s. miller, (mm, agoric):, I think what we do for the for the import.

D

Mark s. miller, (mm, agoric): expression, is that if there's.

D

Mark s. miller, (mm, agoric):, the beginning of a comment.

D

Mark s. miller, (mm, agoric): our yeah possible comment after the import keyword., we already decide.

D

Mark s. miller, (mm, agoric): that it's unsafe. we have to use some trick like that to make sure we don't get confused by the ambiguity. if, when a comment begins and ends.

C

Nicolò ribaudo :: it's a check. if recently confirm that it's like that.

C

Nicolò ribaudo :, let's for, and then either or comment.

D

Mark s. miller, (mm, agoric):, I'm sorry. could couldn't quite hear that.

C

Nicolò ribaudo :, I I I checked with the and the code like a few days ago, like it's exactly as remember it's important for by either of not in parentheses or by a comment.

C

Nicolò ribaudo, : and.

D

Mark s. miller, (mm, agoric): okay,, so I think the same rule would work for, you know.

D

Mark s. miller, (mm, agoric): for dot,, followed by confusing comment,, followed by identifier.

G

Mathieu hofman: yeah,, I think another issue is that our current rule and that's more of a session as of today, problem. but or critical role, would not cache the new in court. Dots source, for example.

G

Mathieu hofman:, I believe. mathieu hofman: and it's been been a while. I haven't looked at it. so.

D

Mark s. miller, (mm, agoric): yeah., so so I believe that about the current long so.

D

Mark s. miller, (mm, agoric): and mark s. miller, (mm, agoric): and in general, we're we're, you know,, biased towards.

D

Mark s. miller, (mm, agoric): not breaking the current rule,, not requiring, says,, to be upgraded in order to remain secure., but you know that's that's.

D

Mark s. miller, (mm, agoric):, we know that there's that., that's a trade off right now, and we're not yet in a position to to. Until we until we have get intrinsics., we know we're not going to be in a position.

D

Mark s. miller, (mm, agoric): to maintain the security of cold cess versions without maintenance.

D

Mark s. miller, (mm, agoric):, once we have get in 10 s., then at some point we'll we'll cross that that bridge,. But we haven't done that yet.

D

Mark s. miller, (mm, agoric):, so mark s. miller, (mm, agoric):, if we really,, if, if the, if the case for doing it with static pseudo properties, is strong,, and I don't know that it's strong.. But if it is strong,, then I think it's,. We can certainly weigh the tray up trade off. with regard to requiring a a security update to to the session.

G

Mathieu hofman:, I I need., I need to go back.. I analyze this a little bit closer. a few months ago. I need to go back to that and see see what I got,, because I I can't.. I just can't remember right. Now.

G

Mark s. miller, (mm, agoric): okay.

B

Guy bedford: yeah, I'd like to just ask again.. So if, if we see.

B

Guy bedford:, I I guess, if we can, it could stop to try to figure out.

B

Guy bedford:, how we we can go about standardizing, monitor, source, and how this will fit into the picture.

B

Guy bedford:, I think that would be a really guy bedford: good thing to to work. On.

B

Guy bedford: obviously., you know.. We we need to get chris's feedback on it., but if we have some plan., then it puts us in a position where, when we're discussing stage 3,.

B

Guy bedford: we're not putting ourselves in a position where there's a risk that we don't get a module source for import reflection, or that it doesn't quite line up.. So I I suppose.

B

Guy bedford:, it would be really bad for sure if you could., I have some.

B

Guy bedford: some way to to figure out module source going forward, or some intensive plan for for doing that.

F

Caridy: yeah., so certainly one it we've got to get was a model there. We we should get.

F

Caridy: as well, the.

F

Caridy: some I'm., I'm still back trying to get back into your previous conversation., still have caps there,. So when, when it comes to.

E

Caridy: getting this source, when you use the source phase.

E

Caridy: you get that source. and in the case of or assembly. you you get that with some of the module.

E

Caridy: which, for all the eventual resource, month, or whatever it is., but once you have that thing.

E

Caridy: is it important to have information about where this.

E

Caridy: source was fetch or not., and I suspect it is because you're gonna do some.

E

Caridy: link it to some point. I don't know.. Is that exactly right, now? and and how that happens, that I don't have it.

E

Caridy: visual indication of what what's going on there?, I don't remember the details.

B

Guy bedford: sure I I.

C

Nicolò ribaudo : yeah. so like in the previous discussions. Also with you regarding,, like specifically components proposals,. I think we settled on the fact that this source doesn't have such info.

C

Nicolò ribaudo :, and that they are all in the like model instance,, which is.

C

Nicolò ribaudo : like what in this wouldn't this live in the and the proposal,. I think, at some point guy even considered having brought him to the.

C

Nicolò ribaudo : for the source phase and import something else for the module it's.

C

Nicolò ribaudo : and,, like me,, that's second.,.

C

Nicolò ribaudo : well,, the third blocking slides, would be where we have that.

C

Nicolò ribaudo : like, for example,, the nicolò ribaudo : and I believe it's not back to the anymore, because, like it's not needed for the.

C

Nicolò ribaudo : motivating this case of this proposal,, which is what some modules.

E

Caridy: okay., so if you,, if I understand that you get the source,, you get something that does not have anything beyond the actual source. Compile, you get the intense you get a source plus the context in which the source was provision., and then you you can do.

E

Caridy: before behavior at that point, and fetch things out of it, and dependencies to to be resolved by the default, hopes, and so on.

E

Caridy:, how? How does that work for a family?, though.

B

Guy bedford:, so for webassembly there is no sense of it. Having a.

B

Guy bedford: any kind of context., so you always provide all of the static imports directly.. So you have to provide for each of the string specifies that it has static imports to., because why, some doesn't have dynamic imports that it's all static.

B

Guy bedford:, you provide the the corresponding instances of those modules. and.

B

Caridy:, but when you do a for instance,, what what do you get for work?, something.

E

Guy bedford:, that's a that's a a a webassembly, dot instance.

C

Nicolò ribaudo : or no like wait,. If you're doing, for instance, of a wasn't model right, now. we'll probably get a module instance object,. That's when imported, like that's nothing, or charles.,.

C

Nicolò ribaudo :, because you cannot import, wasn't models directly., now. however,. If they were platform.

C

Nicolò ribaudo :, like involves the wasn't you an integration that proposes to log.

C

Nicolò ribaudo :, like yes, 7 parts of 1,000, files., that with this importance.

C

Nicolò ribaudo :, you would get a module that represents like a mode of modeling,, some subject that would work.

C

Nicolò ribaudo : when later imported, like when later the.

C

Nicolò ribaudo :, but like right now.

C

Nicolò ribaudo :, like a.

C

Nicolò ribaudo : is meaningless, because there is no.

C

Nicolò ribaudo :, it wasn't nicolò ribaudo, : like.. There is no concept yet that maps.. I was a model to some like context,, based on like it's url, or.

C

Nicolò ribaudo : like where it was imported, from.

E

Caridy: okay,, so does that mean that it will for all.

E

Guy bedford:, while we so we're we're not specifying the instance, import and import reflection.

B

Caridy: and we're we're only currently specifying the source I see.. So in the future we see an extension point where we can. how many times we figured out, okay, got it?.

E

Guy bedford: yeah, so that there can potentially be the extension to that. and once we have the as an integration full as I'm figured out., you could then get the same object that you would for js for the instances,, whereas the sources would maybe be distinct. Objects.

B

Guy bedford: yeah., and then this also brushes up against the csp. discussion,, which is another thing we're following up around the stage 3 process.

B

Guy bedford: and and the.

B

Guy bedford: for for csp., it's a denial at the point of we were investing in this yesterday, at both at the point of.

B

Guy bedford:, when you guy bedford:, first and nicola,, please correct me.. If I got this wrong when you first fetch the module,, so it's actually a compile guard.. So you have to have permission to compile these bytes.

B

Guy bedford: and then, again,, when you evaluate the module, you must have permission to evaluate. and when we integrate with csp. currently,, there is no csp. integration for asm.. So one of the benefits of this proposal is that we can build a csp integration.. That's.

B

Guy bedford: part of the script source, so that you can., you have a rule rule space system to say,. Certain urls are permitted.

B

Guy bedford: to provide was invited for compilation; others aren't;, whereas there currently isn't anything like that for us,, and so that would be able to be gated at the import level,, and that would be a a compilation. Gate.

B

Guy bedford:, it could only be an evaluation gate if we do associate some data with the wise and module. and that's something that we're working out at the web. Assembly, spec: side.

B

Guy bedford: and and yeah, yeah, still under discussion.,.

D

Mark s. miller, (mm, agoric):, so so so there's.

D

Mark s. miller, (mm, agoric):, in considering that the it realizes. it makes me remind me that there's.

D

Mark s. miller, (mm, agoric):, 2 kinds of source module, which would be thinking about from the javascript side., which is the the one that we've been talking. About.

D

Mark s. miller, (mm, agoric):, as well as the virtual source. module the.

D

Mark s. miller, (mm, agoric):, the.

D

Mark s. miller, (mm, agoric):, the source module., that's just author behavior to an api that is not.

D

Mark s. miller, (mm, agoric): derived from mark s. miller, (mm, agoric): compilation of separate sources., and this is already reflected in.

D

Mark s. miller, (mm, agoric): the session support for.

D

Mark s. miller, (mm, agoric): doing sess in environments that limit eval or prohibitive out where you can.

D

Mark s. miller, (mm, agoric):, but directly mark s. miller, (mm, agoric): provide a the the the front door.. That's the sort of the the.

D

Mark s. miller, (mm, agoric): the central behavioral component of asking a module to evaluate with a set of imports and exports.. So the.

D

Mark s. miller, (mm, agoric):, so the the problem,, the reason I'm I'm. well, first of all,, I'm raising this.

D

Mark s. miller, (mm, agoric): generally, because we should keep our eye on it as part of what it means to.

D

Mark s. miller, (mm, agoric): have a notion of source module that that works for the overall epic.. But the other thing is that.

D

Mark s. miller, (mm, agoric):, there's no conflict with regard to csp. inhibition of compilation in the first place, and csp. inhibition.

D

Mark s. miller, (mm, agoric): of mark s. miller, (mm, agoric):, whether sources from a particular place, are compiled to produce a source module., but once you've got the source module.. The issue of.

D

Mark s. miller, (mm, agoric):, is it somehow mark s. miller, (mm, agoric): unforgivably tagged with origin, information.

D

Mark s. miller, (mm, agoric):, so that you can then do further csp. restrictions on it.

D

Mark s. miller, (mm, agoric): well,, there's.

D

Mark s. miller, (mm, agoric):, the notion of some kind of unforgeable tagging is at odds with.

D

Mark s. miller, (mm, agoric):, the notion of.

D

Mark s. miller, (mm, agoric): our virtual source modules that that that it's simply an api, that you can implement directly with that.

D

Mark s. miller, (mm, agoric):, without having compiled it from sources,, and it can act like.

D

Mark s. miller, (mm, agoric):, like a source model., the other thing, of course,, is that there are no origins in the tc. 39 spec.

D

Mark s. miller, (mm, agoric): so mark s. miller, (mm, agoric):, so let me just pose the question., so the.

D

Mark s. miller, (mm, agoric): is there actually a need for csp. controls, post compilation?.

D

Mark s. miller, (mm, agoric):, you know what you can do with a.

D

Mark s. miller, (mm, agoric):, you know a source module, once it exists.

D

Mark s. miller, (mm, agoric):, where are the purposes of the csp eval restrictions,, all served by preventing the compilation into a source. module. in the first place.

B

Guy bedford: I'm. having a meeting with with francis, who it worked on the csp specification for.

B

Guy bedford:, hopefully next week,, and I would like to follow up on this and report back.

B

Guy bedford:, some of those questions for us, and and how those have been understood in in the wisdom, ecosystem.

B

Guy bedford: nicola,, you had your hand up briefly.. There was there anything that I got.

C

Nicolò ribaudo : no,. I just want to answer to mark.

C

Nicolò ribaudo :, so nicolò ribaudo :, like there is at least one difference between like built in services and your 12 sources.

C

Nicolò ribaudo :, because nicolò ribaudo : built in services can be like potentially transferred across agents, or like across,, like well, beautiful sources cannot, because they like they would probably contain.

E

Caridy: who they are. on saying that I'm I'm not. at this point, I'm not very convinced that we need to withdraw more with our sources.

E

Caridy:, we might be able to get away without that.

D

Mark s. miller, (mm, agoric):, so right now we are successful. using virtual sources in the session in particular,, as used by metamas.. In order to.

D

Mark s. miller, (mm, agoric): in order to turn a common js. modules into module, sources.

E

Caridy: right?, but since the source does not doesn't connect to another source directly, it goes through an intense.. You might very well do the same thing at the intense level.

E

Caridy:, so imagine that, rather than creating a and you.

E

Caridy: on a source that you import, somehow, or that you create using the module source. constructor.. Imagine that you can create a a module intense that does not have a source, but does have the controls that allow you to define.

E

Caridy:, how the name and space of it is going to be built.

D

Mark s. miller, (mm, agoric):, I mark s. miller, (mm, agoric):. I mean a module instance already.

D

Mark s. miller, (mm, agoric):- I I don't understand that the the essential thing about a module source is that it can be instantiated.

D

Mark s. miller, (mm, agoric): separately in in different compartments,, with full control.

D

Mark s. miller, (mm, agoric): by the compartment over.

D

Mark s. miller, (mm, agoric):, you know how it gets, how it gets instantiated, and what it gets hooked up, to, and what the important namespace.

D

Mark s. miller, (mm, agoric):, you know is.

D

mark s. miller, (mm, agoric):. We can do that with the common js modules by virtue of turning them into virtual virtual source modules.. If you turn them directly into module instances,, you would have to turn them.

D

Mark s. miller, (mm, agoric): directly into module instances separately in each compartment, because there's commitments that are made when turning it into a module instance that are not made, and turning into a module to source.

E

Caridy: right?, but that should not be.. That should not be a a deal record. You you!, you can still create factors of these module instances per component.

D

Mark s. miller, (mm, agoric):, when you so mark s. miller, (mm, agoric):, how does it fit into the compartment? api., I don't., I don't understand.. How does it fit this factory of instances?? What is the factory of instances, if it's not a virtual.

D

Caridy: module source., so I I think the key, the key.. The key point is that.

E

Caridy: a source does not connect one more source.. You need a vehicle to connect sources to get them link and creating instances out of them and link them, together, and so on.. So the source,, using it as a vehicle only to share certain information that actually belong to the same room., but you're. Sharing that information.

E

Caridy: across different compartments.

E

caridy:, and with that information. you're able to go ahead and create instances.

E

Caridy: that allow you to that that has a certain mechanism to link.

E

Caridy: different instances together and so on.

D

Mark s. miller, (mm, agoric): okay.. But what is the reification of the.

D

Mark s. miller, (mm, agoric):, the information provided only by the source?.

D

Mark s. miller, (mm, agoric): such that you can feed that into the compartment? api. and.

D

Mark s. miller, (mm,, agoric): and instantiated differently in different compartments.

E

Caridy: right?, so they only the only detail. There is that you, you have to have.

E

Caridy: knowledge about what what kind of interest you are creating when you have the.

E

Caridy: virtual source all sometime, and do not necessarily need that additional information.. You just do a creation of an intense.

E

Caridy: off of a source, whether that's the source that you import or source that you build, use individual source.

E

Caridy: and that is nicer. caridy:, but my argument there will be that you still need to have significant knowledge about what this is, because you have to provide.. I'm sorry, I'm I'm getting lost in the pronouns. who has to have knowledge.

D

Caridy:, the whoever is creating an intense out of a source.

E

Caridy:, whether that's virtual source or or a input source,, because you need to do the the the the linkage of that intense., you need to have information about what you're creating.. What is the thing that you're creating,, so you can build the right.

E

Caridy:, whatever form of linkage, you will use.

E

Caridy: I mean the the idea,, I mean the division of responsibility,. As I understand it.

D

Mark s. miller, (mm, agoric):, is that mark s. miller, (mm, agoric):, the the common js virtual,. You know the author of the common js.

D

Mark s. miller, (mm, agoric): virtual module mechanism. the thing that turns a common js. source into a common js virtual source module.. He has to know that it's obviously that it's a common js monitor., but once.

D

Mark s. miller, (mm, agoric):, the virtual source, module is created.

D

Mark s. miller, (mm, agoric):, then somebody using the compartment api.

D

Mark s. miller, (mm, agoric):, doesn't particularly have to know that that virtual source module, represents.

D

Mark s. miller, (mm, agoric): I'm. in js. it just. it has imports and exports and gets hooked together.. You know, there's certain restrictions that you know certain pain points of the.

D

Mark s. miller, (mm, agoric): how come, you know. fits in., but but except for edge conditions,. Somebody using the compartment api to to instantiate and link.

D

Caridy:, these virtual source modules does not need to know that their common js modules. yeah., so that that that is precisely the point I'm trying to make. It doesn't need to know that it's a common,, but it needs to know what it is,. How how to link to other things.

E

Caridy: but I mean it's it's the way it links to other things.

D

Caridy: is,, except for edge conditions,, exactly the same way., any other source. module links to other things.

E

Caridy: or not.

E

caridy:, I give it to you caridy: to build the intens it out of them without knowing what they are.

D

Caridy:, when you say,, knowing what they are,, I need to know what they import next board.. I don't need to know what their source language is.. So you you do need to know,, because if they have,, if they have a dependency., how do you resolve the dependency without knowing what that thing, is? because, remember: the source doesn't have.

E

Caridy:, a origin of any kind.

E

Caridy:, it's just a source.

E

Caridy:, so I give 2 of them to you.. What do you do with them? to resolve to the input that they have?? Both of them has an import. x.

E

Caridy: got you inside that.

E

nicolò, ribaudo :. How do you know how to resolve that there?? How is this different from the team services like?? Even if I have a javascript to js.

C

Nicolò ribaudo :, you still need to like have,, and that is precise in my point.. What I'm trying to say what the the the gist of it is that.

E

Caridy:, the source itself is not sufficient.

E

Caridy:, even if we introduce a virtual source that is not sufficient., you have to know what you're doing.

D

Mark s. miller, (mm, agoric):, you have to know what the what the purpose of the module is., but I mean if, if the module was one day written in common js. and another way., another day, written in ecosystem modules.

D

Mark s. miller, (mm, agoric): that refactor should be mostly invisible to compartment to you know, code at the compartment level, that's manipulating it to to instantiate, to link and instantiate it.

E

Caridy: certainly,, but because you already have to have some sort of out of.

E

Caridy: configuration for the thing that you're going to integrate you. could.. You could use that out of one sign all.

E

Caridy: to create instances that are tailor made to the source that you want, rather than having to create a virtual source.. So the argument that I'm trying to link is again. but but, as you say, create the instances,. The creating.

D

Mark s. miller, (mm, agoric):, I I I'm really not understanding this,. The creating of the instances is done by using the compartment. api., the the what?. What's that into the compartment? api. are module sources of virtual module, sources.

D

Mark s. miller, (mm, agoric):, so the mark s. miller, (mm, agoric):, the the the person who knows that.

D

Mark s. miller, (mm, agoric):, this module source represents a common js module was not the one using the compartment api.. It's only the one.

D

Mark s. miller, (mm, agoric):, who mark s. miller (mm, agoric):, created the mechanism for turning common js sources into a virtual source. module.

D

Mark s. miller, (mm, agoric):, because if it gets refactored.

D

mark s. miller, (mm, agoric):, it's only the replacement of that mechanism. That happens when it's refactored to change its language.

D

Mark s. miller, (mm, agoric): the the code, that's using the compartment api to instantiate link.. It should not need to change.

F

E

Caridy: trying to find a different way to explain my mental model. here.

D

Caridy:, I see somebody has their hand up.. Is there a question that we're not getting to.

E

Caridy:, I I feel that the.

E

Caridy:, when, when it comes to virtualization or.

E

Caridy:, so it feels to me that we will. what you're trying to achieve is that you do virtualization by using the concept of.

E

Caridy: a so as the vehicle to achieve utilization.

E

Caridy:, therefore, there, the code running inside the the the compartment doesn't need to know that you're getting a source or a virtual source.. We just use that thing,, assuming that it's a sort., but it happens to be a virtual source,, because the virtualization.

E

Caridy: or the hopes that we have on the compartment level allows to replace that, with the virtual version of what you're trying to import.

D

Mark s. miller, (mm, agoric): yeah,, that's that was. yes, exactly.. That's that's my understanding. What the what the whole point of virtual source modules, is. yes,! What I'm trying to argue is that you should be able to achieve.

E

Caridy: the same thing without having him to rely on a virtual source.

E

Caridy:, you can just use the source.

E

caridy:, who can you get the same separation of concerns.

E

Caridy:, I believe so,, because what you're gonna give to them is basically a source whose job is to import.

E

Caridy: running module that you will resolve.

E

caridy: and this module is going to be.. We can call our virtual module,, but it's not a it's, actually a module. That does not have a source,, but it has the control that allow you to create the thing that you need. so basically you,. That is a little bit more cumbersome because you have an intermediate module. Now.

E

Caridy:, but it's it's an export.. I I expert.

D

Mark s. miller, (mm, agoric): oh,, an intermediate module., I okay. now I'm starting to see., but you still,. That means you have to write an intermediate module per.

D

Mark s. miller, (mm, agoric): foreign language, module that you're trying to virtualize,, basically for every virtual, for every virtual module you wanted to create., you have to create an actual intermediate module.

D

Caridy: is that what you're saying?? What if at at the end of the day, the visualization is simply going to resolve to a source of some sort, or whether that's the source of now,? What does it know existing in terms of a module that you want to kind of.

E

Caridy: proxy to it, or something like that,, you still be able to create an intermediate source. that that's what you give to.

E

Caridy:, the the actual compartment., so they get a source just like they were getting before they get a source.. It's just that they source. the only thing that it does is re-export everything from another module that you will result. To.

E

Caridy: and you provision that model somehow, in such a way that this module that you are going to provision is going to define via some api.. What is the, what is it that is?? It is exporting.

E

Caridy:, rather than using a a they sort as a vehicle to define,, they.

E

Caridy: export and the import from, or the modules.

D

Mark s. miller, (mm, agoric): okay,, let me.! Let me first of all admit that I'm only half getting this clearly., I think we need to follow, follow up offline to arrive at a mutual understanding., and I apologize that I ran out the time on this one issue.

D

Mark s. miller, (mm, agoric):, when we have other issues that we should have also gotten to.

D

Mark s. miller, (mm, agoric): and of course, that we also did not get to any of the tc 39 stuff.

D

Mark s. miller, (mm, agoric):, we do have a next meeting that we uses the same zoom room. so.

D

Mark s. miller, (mm, agoric):, it's not possible to just let this meeting continue past the the hour.

E

Caridy: yeah,: let's, let's shout off line., I think they they only the television.. We we. We should probably not all that strong onto the virtual module.

E

Caridy: for preventing, or things to be explored.

E

Caridy:, because it might be a way that we can just do not need to have a promote auto., I we figure that out, and then.

E

Caridy: the the concerns that we have related to virtual module, not really concerns.

D

Mark s. miller, (mm, agoric): okay., so let's follow up offline.

B

Guy bedford:, I would like to follow up on this csp discussion next week.. Maybe if, if that makes sense,, I can create an item for that. and yeah,. Sorry to take up all the time on on this topic.

B

Guy bedford:, I guess maybe if we can make it short of next week, there could be time for wider tc: 39 discussion. as well.

D

Mark s. miller, (mm, agoric): okay, a guy., thank you very much. and yes,. I think that bringing bring this topic, and and with the focus on the csp issue next meeting would be great.

C

Nicolò ribaudo : okay,, so thanks everyone and see you at some point in the future.

C

Nicolò ribaudo : bye.

A

Mark s. miller, (mm, agoric): bye.