| 13:33 | <devsnek> | is anyone else super confused that realms are on the agenda for stage 3 |
| 13:35 | <ryzokuken> | ...because it's been years since they were Stage 2? |
| 13:39 | <devsnek> | no cuz it doesn't appear that they solved the issue brought up last time it was presented |
| 13:47 | <littledan> | devsnek: What do you see as the important unresolved issues? |
| 13:47 | <littledan> | I think the slide deck proposes resolutions to everything |
| 13:48 | <devsnek> | call boundary |
| 13:48 | <littledan> | we frequently have a pattern in TC39 where people raise directly contradictory points and the resolution is, "choose one over the other" |
| 13:48 | <devsnek> | i don't think this proposal solves the criteria set out for it with the call boundary |
| 13:48 | <littledan> | which criteria are unsolved by the current spec text? |
| 13:49 | <devsnek> | basically all of this https://gc.gy/89655536.png |
| 13:49 | <devsnek> | all of the examples in your explainer break of any of the functions return an object lol |
| 13:50 | <littledan> | I don't think it's very polite to lol at proposals |
| 13:50 | <littledan> | these examples can work through the use of membranes or other indirect techniques |
| 13:50 | <devsnek> | i might expect the proposal to include membranes or other indirect techniques in that case |
| 13:51 | <littledan> | well, membranes are excluded because membranes are more of a pattern that can be taken in many different directions, rather than something which only has one implementation |
| 13:51 | <littledan> | the champions have provided a membrane library that can be used with the proposal |
| 13:52 | <devsnek> | if objects make the api throw, there needs to be a built-in way to deal with that, is what i mean |
| 13:52 | <devsnek> | if recursive proxy magic is the answer, i'd want an api to create recursive proxy magic |
| 13:56 | <devsnek> | littledan: i think new Realm({ unsafeSandboxWillBeBrokenAllowObjects: true }) would be fine too |
| 14:01 | <littledan> | OK, it's true that this point was raised last time, and I just disagree with it |
| 14:02 | <littledan> | you can make internal slots work by always using an object with the internal slots that was allocated from the outside |
| 14:03 | <devsnek> | i'm just not sure how i would use this proposal for anything in its current state |
| 14:05 | <devsnek> | and the importValue api seems like a footgun considering it might give you an object or a function which returns an object |
| 14:13 | <devsnek> | littledan: do you have a link to the membrane library that can be used with this proposal? i can't find it from the repo |
| 14:17 | <littledan> | I really don't understand this concept of a footgun. It feels like everything under the sun gets accused of being a footgun |
| 14:17 | <devsnek> | i would always use evaluate instead of importValue |
| 14:17 | <devsnek> | cuz with evaluate you can wrap the function to catch its return value |
| 14:19 | <devsnek> | also proxies can't get through the call boundary |
| 14:19 | <devsnek> | do i have to pass lists of object keys as comma separated string values |
| 14:21 | <devsnek> | i'm just trying to imagine how i'd port my cloudflare worker test thing from node vm to this |
| 14:29 | <littledan> | well, evaluate and importValue both do the exact same function wrapping |
| 14:29 | <littledan> | it's just that they differ in terms of sync/async and CSP |
| 14:29 | <devsnek> | no i mean i would do function wrap(f) { return (...args) => { f(...args); } } |
| 14:29 | <devsnek> | no return value at all |
| 14:30 | <littledan> | OK, sure |
| 14:30 | <littledan> | I mean, depends on what you're trying to do |
| 14:30 | <littledan> | you can also do ,0 of course |
| 14:30 | <devsnek> | i'm not trying to sandbox bad code |
| 14:30 | <devsnek> | i'm just trying to virtualize |
| 14:31 | <littledan> | well, this Realm proposal is simply less expressive than the Node vm module |
| 14:31 | <littledan> | it doesn't aim to do everything; it aims to meet the goal of providing some kind of boundary with a little bit of integrity and separation (though not Spectre-level) |
| 14:32 | <devsnek> | sure but in that process i'm not sure how you do anything with it |
| 14:32 | <devsnek> | like one of the the examples in the explainer adds two numbers together |
| 14:32 | <devsnek> | very cool but now what if that function does something more complex |
| 14:33 | <littledan> | well, I'll find the membrane example for you |
| 14:33 | <devsnek> | also how do membranes work with importValue |
| 14:33 | <devsnek> | as far as i can tell there's no hook point there |
| 14:34 | <littledan> | no, but there's no hook point with evaluate either... |
| 14:34 | <devsnek> | with evaluate you insert makeMembrane(v) or something right? |
| 14:35 | <littledan> | you can also replace importValue with something which does the I/O and transpiles the module system (this is how all module system virtualization works today anyway) |
| 14:35 | <littledan> | or, if you're in Node, sure, you can hook it |
| 14:36 | <littledan> | I think it's quite important that we have a version of evaluating code in Realms which doesn't violate strict CSP policies |
| 14:36 | <devsnek> | i'm just thinking like, import handleRequest from a module, pass a complex request object to it, and use the complex return value |
| 14:36 | <devsnek> | curious to see how much abstraction i need for that |
| 14:37 | <littledan> | I don't understand what you mean or what you're trying to do |
| 14:37 | <devsnek> | mocking a cloudflare worker environment |
| 14:38 | <devsnek> | get an http request from node's http stack, turn it into something the cf worker understands, pass it into the worker code which is inside the realm, get the return value out, turn it into a response node understands |
| 14:39 | <devsnek> | seems like a fairly standard use case for virtualization |
| 14:41 | <devsnek> | with realms api, you could even make a version that runs in the browser |
| 14:41 | <yulia> | I am a bit lost, can you sketch out how you think this would work in realms? I thought it was pretty straight forward how you get the result out of a realm |
| 14:41 | <littledan> | yeah, basically with the callable boundary realms proposal, you have to serialize all of these operations in terms of primitives |
| 14:42 | <devsnek> | yulia: a cloudflare worker is a module which exports some functions. in the old realm proposal i think it would've looked something like const { handleRequest } = await realm.import('./cf-worker-cide') and then const r = fromFetchResponse(handleRequest(toFetchRequest(node request))) |
| 14:43 | <littledan> | right, you can't just share objects synchronously with this callable boundary realm proposal, but you can build a membrane out of it, see https://github.com/caridy/irealm |
| 14:43 | <littledan> | (and it turns out that this doesn't need symbol as weakmap key) |
| 14:44 | <devsnek> | so i can import irealm and get something more like the old proposal? |
| 14:44 | <littledan> | with this support library, yes |
| 14:44 | <devsnek> | i wonder what the perf overhead is 😰 |
| 14:45 | <littledan> | well, that seems like something we can examine during Stage 3? |
| 14:45 | <littledan> | also there may be room for improvement in Proxy performance |
| 14:45 | <devsnek> | sure... |
| 14:46 | <devsnek> | i would probably just keep using node vm |
| 14:47 | <devsnek> | i'm just not sure why this is better than adding an "allow objects" option |
| 14:47 | <littledan> | it's fine with me if you keep using node vm |
| 14:47 | <littledan> | well, the presentation explains the motivation for the callable boundaries approach |
| 14:47 | <devsnek> | or for that matter moving GetWrappedValue to userland |
| 14:47 | <devsnek> | since afaict it doesn't do anything js can't do? |
| 14:48 | <littledan> | OK, I believe this has already been explained in several threads and presentations that you've read |
| 14:48 | <littledan> | so I'm not sure what I could say beyond repeating that content |
| 14:49 | <devsnek> | aight |
| 14:49 | <devsnek> | my point is just, if this is unfulfilling to the point of a case like mine not using it, it doesn't seem to match the problem the committee wanted to solve |
| 14:50 | <littledan> | well, I explained how it can fulfill your use case |
| 14:50 | <devsnek> | sure, but that's a gamble on my part |
| 14:51 | <devsnek> | if the perf can't be solved, i'm stuck being the asshole who blocks stage 4 |
| 14:51 | <littledan> | because of performance? |
| 14:51 | <littledan> | I guess I'd be OK with optimizability being a Stage 4 blocker, personally |
| 14:52 | <littledan> | much of the motivation for this proposal is that it is more optimizable than creating whole iframes |
| 14:53 | <devsnek> | rereading some stuff, i didn't see anything explicitly explaining why an option on the constructor is unacceptable |
| 14:53 | <devsnek> | could've missed it |
| 14:55 | <littledan> | well, the argument against that option would be the same as the argument for callable boundaries in the first place |
| 14:55 | <littledan> | namely that it's a sort of integrity/boundary footgun |
| 14:56 | <devsnek> | i mean if the alternative is pushing people to use iframes or node vm |
| 14:56 | <devsnek> | lol |
| 14:58 | <littledan> | well, yeah, I think we will push people to iframes or node vm if we can't agree on callable boundary, since it will make it hard for the proposal to move forward at all |
| 14:58 | <devsnek> | like as one of the maintainers of node vm, i was really hoping to be able to deprecate it |
| 14:59 | <devsnek> | because it is awful |
| 15:00 | <littledan> | Node vm is really much more expressive than even the non-callable-boundaries Realm proposal |
| 15:00 | <littledan> | so it's not clear to me what it would mean to deprecate it |
| 15:01 | <devsnek> | well compartments are supposed to provide host hooks right? |
| 15:01 | <devsnek> | with those two things we could kill it |
| 15:02 | <littledan> | yes, that is possible. Similarly, sharing objects directly as a Realm constructor flag could also be a follow-on proposal |
| 15:03 | <devsnek> | i mean it sounds like you'd be against that |
| 15:04 | <littledan> | I'm not personally against it, but I understand the arguments against it |
| 15:04 | <littledan> | I don't know if we'll ever get consensus on either compartments or shared-object Realms |
| 15:04 | <littledan> | to go to Stage 3 on this, we have to be OK with the potential future where they never happen |
| 15:22 | <devsnek> | littledan: one more question i guess... what about irealm prevents it from being something we provide in stdlib |
| 15:22 | <devsnek> | is there some weird opinionated choice it makes? |
| 15:24 | <littledan> | well, maybe it could eventually be in the stdlib; I wouldn't be opposed to that, but I also don't see why we should block on it. |
| 15:25 | <devsnek> | it seems like a reasonable alternative to throw a TypeError but 🤷♂️ |
| 15:35 | <littledan> | it adds a lot of complexity. I'm OK with starting with something simpler |
| 17:17 | <ljharb> | (i can't post in the delegates channel btw) |
| 17:17 | <ljharb> | why is matrix hijacking the Mac OS's default command-` keyboard shortcut to switch windows? can i turn that off? |
| 17:17 | <jridgewell> | Same |
| 17:18 | <ryzokuken> | why is matrix hijacking the Mac OS's default command-` keyboard shortcut to switch windows? can i turn that off? |
| 17:18 | <ljharb> | no, that's to switch apps. command-tilde switches windows within the same app |
| 17:19 | <ryzokuken> | ah, I see. Let me see why it's doing that. |
| 17:19 | <Tierney Cyren> | how do I join a space? |
| 17:19 | <Tierney Cyren> | or spaces or whatever it's called |
| 17:19 | <ryzokuken> | how do I join a space? |
| 17:19 | <ryzokuken> | spaces are just rooms with special flags set |
| 17:19 | <ljharb> | is there no threading, it's just quoting replies? |
| 17:20 | <ryzokuken> | yeah, no threading yet, no |
| 17:29 | <shu> | can someone give me permission to send messages in the delegates channel |
| 17:30 | <Tierney Cyren> | cc @ryz |
| 17:30 | <ryzokuken> | tab |
| 17:30 | <ryzokuken> | 😛 |
| 17:30 | <Tierney Cyren> | why didn't that autocomplete |
| 17:30 | <ryzokuken> | you need to hit tab manually |
| 17:30 | <ryzokuken> | sadly, I don't have access. |
| 17:30 | <ryzokuken> | cc littledan |
| 17:33 | <SoftwareChris> | can someone give me permission to send messages in the delegates channel |
| 17:46 | <yulia> | ljharb: we can't hear you |
| 17:46 | <ljharb> | hm, ok thanks |
| 17:47 | <ljharb> | brb |
| 17:49 | <yulia> | Issue and answer ljharb https://github.com/bocoup/test262-report-issue-tracker/issues/27#issuecomment-840203670 |
| 17:51 | <bakkot> | does anyone know how to make the interface on element.io less noisy, or am I going to download a desktop client? I would like to be able to see more than like five lines of the chat at a time |
| 17:52 | <Tierney Cyren> | zoom out? |
| 17:52 | <yulia> | bakkot: do you mean, the spacing? |
| 17:52 | <ljharb> | ooh, i enabled "irc style layout" and it's much nicer and more compact (@bakkot) |
| 17:52 | <bakkot> | ljharb: nice! where's that? |
| 17:53 | <ljharb> | in "advanced" appearance settings |
| 17:53 | <ryzokuken> | settings > appearance |
| 17:53 | <bakkot> | yulia: yeah, I mean the the whitespace to actual content ratio |
| 17:53 | <Michael Ficarra> | @bak |
| 17:53 | <Michael Ficarra> | bakkot: the settings |
| 17:54 | <yulia> | go to this thing |
| 17:54 | <yulia> | click "show advanced" |
| 17:54 | <bakkot> | sweet, thanks |
| 17:54 | <yulia> | ^ that |
| 17:54 | <bakkot> | I actually had not managed to find the settings yet |
| 17:54 | <bakkot> | they were under my username, which is not where I was expecting to find them |
| 17:55 | <littledan> | ooh, i enabled "irc style layout" and it's much nicer and more compact (@bakkot) |
| 18:28 | <yulia> | shu: i think we are still waiting on editor signoff is that right? |
| 18:32 | <shu> | yulia: for TLA spec? |
| 18:33 | <yulia> | yeah |
| 18:34 | <shu> | yes, still needs editor sign off before mergeable |
| 18:34 | <shu> | i need to work through your example still |
| 18:34 | <shu> | in general the TLA spec is just very difficult to understand :( |
| 18:37 | <yulia> | yeah :( i am happy to pair on it if it helps |
| 18:38 | <yulia> | i would like to refactor the module loading spec from loops to recursive calls? i think that would help? |
| 18:38 | <yulia> | like, eventually, as an editorial change |
| 18:39 | <shu> | i'm tempted to not change that stuff because we'd need to review the recursive versions again |
| 18:39 | <shu> | i'm also not excited about chasing down all the outdated spec comments in the implementation |
| 18:42 | <yulia> | fair |
| 18:42 | <yulia> | same |
| 21:03 | <Richard Gibson> | can someone grant me permission to post in Delegates? |
| 21:04 | <ryzokuken> | Aki: Robin Ricard Rob Palmer ^ |
| 21:05 | <Robin Ricard> | someone should remove me admin rights |
| 21:05 | <Robin Ricard> | or can I do it to myself? |
| 21:06 | <Robin Ricard> | I removed my own admin rights |
| 22:08 | <shu> | someone should remove me admin rights |
| 22:47 | <Aki> | this is also how i aspire to live. please, unburden me of responsibility |
| 23:52 | <shu> | don't get me wrong, i'm not touting my moral character here |
| 23:52 | <shu> | i'd take authority without responsibility |
