freenode :: #whatwg

18 Apr 2017
00:16DomenicWell, I manage to crash chrome when pushState()/replaceState()ing shared array buffers :D
04:01annevk
04:05DomenicServo is implementing MathML? O_O
04:36annevkI suspect it's not Mozilla driving that
04:37annevks/suspect/speculate/
04:37DomenicManishearth is doing the work
04:37DomenicHmm, it's for Stylo, I guess the idea is to not regress if switching Gecko to use Servo components.
04:38annevkOoh, that makes sense
04:39annevkI still like that MathML became a thing of sorts; kinda sad it's still lingering
04:46MikeSmithr? https://github.com/whatwg/wattsi/pull/43
04:49ManishearthDomenic: I'm implementing mathml in stylo, yeah :)
04:50Manishearthwe don't want to unship things alongside stylo
05:19TabAtkinsUgh, I finally found a justification for &quot;record<> as return value&quot;.
05:20DomenicHmm, I&#39;m curious what its advantage would be over object. I guess if you already have a record<>.
05:23TabAtkinsI mean, same as `object`, whatever. `record<>` is just an `object` that advertises its types better.
05:23TabAtkinsBut I mean vs returning an interface object, which I previously always recommended.
05:24TabAtkinsContext here is that some of the TypedOM stuff we decided today will let you reduce a calc() expression into a simple sum of values; we&#39;ll return this as an object with units for keys and doubles for values.
05:31MikeSmithTabAtkins: gentle reminder to please nudge plinss about the <{link/rel/serviceworker}> thing (if hes there yet)
05:32TabAtkinsYeah, he&#39;s in. We&#39;re all in the Houdini meeting right now.
05:34MikeSmithk
05:36annevkDomenic: given our discussion yesterday, should we really use Data Block in serialize?
05:36annevkDomenic: it seems we should serialize that too, otherwise storage becomes kinda weird, keeping the ECMAScript environment alive to support that structure
06:08MikeSmithDomenic: r? https://github.com/whatwg/wattsi/pull/44
14:20Domenicannevk: I don&#39;t see the problem with Data Block. It&#39;s bytes in memory; it isn&#39;t associated with the ECMAScript environment as far as I can tell.
14:22annevkDomenic: yeah, I guess it&#39;s just Shared Data Block
14:22DomenicIs a SDB associated with an ECMAScript environment?
14:22annevkDomenic: with an agent
14:22annevkDomenic: or agent cluster rather
14:23DomenicRight OK
14:23DomenicI don&#39;t think it keeps the agent cluster alive though
14:23annevk(although I think this is currently not explicit awb-style)
14:23DomenicI think of it as more having access checks that if the wrong process tries to access its bytes, the browser crashes
14:23DomenicNot a pointer to an agent cluster
14:23DomenicMaybe a GUID identifying the agent cluster
14:28annevkDomenic: will probably fix foolip&#39;s comments in https://github.com/whatwg/html/pull/2521 tomorrow; did the US have a holiday yesterday too or is it likely that getting further comments on that PR is going to be hard without more pushing?
14:28annevkDomenic: or maybe I should reset my expectations and wait a week?
14:30Domenicannevk: no public holiday; it does seem people are slow. My take is we can wait until we&#39;ve lined up all the other PRs, including tests, one for service worker, and if we&#39;re ambitious the StructuredSerializeForStorage switch-over PRs. If nobody&#39;s commented by then we should go ahead.
14:31annevkDomenic: I guess we could also wait and see if your doom scenario of dropping agent clusters happens
14:31DomenicI think they&#39;re still useful as a memory boundary even if they don&#39;t have all the properties lth wants
14:34annevkDomenic: okay, so I&#39;ll fix up the comments and then once all PRs are reviewed again I&#39;ll rebase and clean up commit messages (there&#39;s some rebasing trouble due to other smaller changes I think)
14:35DomenicSounds good. I&#39;ll keep working on tests and maybe work on a service worker PR.
14:35annevkDomenic: I&#39;m happy to do the PR for Notifications and I was also think of switching DOM to using an agent rather than a unit of browsing contexts
14:35Domenicah interesting
14:36annevkThe other thing I plan on working on is Location/Navigator, probably with proper slots
14:36annevkBut I need to write some tests first to make sure they&#39;re actually associated with the global and not the document
16:52annevkDomenic: did Apple participate in the WebAssembly call?
16:52Domenicannevk: yep
16:52annevkDomenic: did you bring up the concurrency issue?
16:53annevkDomenic: well, independent worker shutdown
16:53Domenicannevk: I did; apparently there&#39;s internal debate and it&#39;s on their radar, which I didn&#39;t know before the call
16:53annevkDomenic: great!
16:54annevkDomenic: as a small aside, there will always be fixup comments
16:54annevkDomenic: and in fact, I&#39;ve already planned a complete rewrite of how we do agents once we figure out where exactly we want to create agent clusters and agents
16:55Domenichaha, fair
19:31annevkDomenic: so I run git blame and wait for a long time (old MacBook Air) to find out about these worker algorithms and I get https://github.com/whatwg/html/commit/cece923873b2947a00d542082f0b50af2c85b0bb
19:32annevkDomenic: so now I&#39;m waiting again
19:34DomenicI usually search commit messages but it might not work in this case
19:36annevkDomenic: from https://github.com/whatwg/html/commit/a2cf6570bab4e42fcac0d576d786b7ce389d117a it seems that specific line predates integration into HTML
19:36annevkI usually use git log --grep=, but that doesn&#39;t work for specific steps really
19:37annevkThough blame is pretty useless too it turns out due to refactoring
19:40DomenicI think the internet has some recursive blame helper scripts
19:40annevkDomenic: I found https://lists.w3.org/Archives/Public/public-html-diffs/2008Dec/0048.html
19:41annevkDomenic: by running svn log https://svn.whatwg.org/webworkers/
19:41annevkDomenic: and then a Google search
19:41annevkDomenic: it very much looks like the divergence is not necessarily intentional
19:41Domenic&quot;who were you ap? what did you see!?!?&quot; https://xkcd.com/979/
19:41Domenic(Wow IRCCloud has built-in xkcd inlineing O_o)
19:42annevkOh, not corporate IRCCloud
19:42annevkI know who ap is
19:43DomenicWell, not intentional sounds believable. Let&#39;s merge.
19:43annevkNick of an Apple WebKit engineer
19:43annevkGave lots of XMLHttpRequest feedback in the early days
19 Apr 2017
No messages
   
Last message: 127 days and 11 hours ago