io.js TC Meeting 2015-03-18

Agenda

Ben: reviewing bugs, fixing pull requests
Bert: debugging Windows issues (related to #1005), working on a more helpful CI dashboard
Chris: working on tools to inspect amount of ecosystem breakage that changes cause
Fedor: working through PayPal leak issue, working on net.socket() changes
Rod: ..
Trevor: misc PRs, process.nextTick(), timer bug

lib: fix process.send() sync i/o regression #774 / @bnoordhuis
iojs: introduce internal modules #848 / @chrisdickinson / @trevnorris / (@vkurchatkin)
url: significantly improve the performance of the url module #933 / @mikeal
Ship websockets in core #1010 / @piscisaureus
RFC: upgrade to V8 4.2? #1026 / @bnoordhuis
Colin resigning from TC #1056

Rod: originally conceived to be a “sub-WG” of Build WG but this makes sense because there is little overlap with the Build team or skills required for the Build team.
Rod: this group is already maintaining official Docker images or io.js located at https://github.com/nodejs/docker-iojs which are also the “official” Docker images for io.js.

Bert: in favour of stream_base now and don’t want it reverted given the fixes that have been made. The real problem was the introduction of a major change that lead to a lot of breakage and a lot of fix commits. Proposed that we have a “ng” (or similar) branch with releases for testing before merging.
Chris: agree that we need a canary policy of some kind to deal with this
Bert: feature flags are a good way of dealing with this and letting them test changes
Ben: time-based releases like Rust and Chrome
Fedor: main problem is that we didn’t have enough tests for this feature when it was merged.
Rod: there’s awkwardness around releasing given that we only have one tip and we can’t cut patch releases once we have semver-minor changes are merged.
Trevor: should have semver-minor branches where patch fixes go in and minor changes go in to master/whatever.
Discussion about git process
Discussion about semver
Discussion about minor/patch and what it communicates
Discussion about test coverage
Rod asked OpenSSL fix: agreed to punt on a release until OpenSSL fix has been released and go immediately after that.

Rod: there was some off-record discussion about TC membership process related to #1130 and #1131.
- Neither individual was asked and it would be good to at least have it passed by them first
- Would be good to have a smooth on-boarding mechanism to bringing someone on to meetings before officially making them TC, agreed to invite Jeremiah to next TC if he’s interested in being involved.
- Generally agreed that increasing the size of the TC is a good thing (minus the meeting-overhead added by more people)

Trevor: passing arguments / context through process.nextTick() in core is common and would be significantly faster if we use a setArgs() feature.
Chris: needed in readable-stream
Discussed API concerns and how underscored methods still become defacto in userland
Chris: could land the internal-module PR and use a Symbol for setArg
Trevor: major speed increase comes from flattening the call-structure, there’s no additional stack entries for passing arguments. Particularly in crypto, nextTick() is called a lot.
Rod: we need perf numbers from a user-experience perspective, perhaps that will make it more compelling to merge or perhaps it will make this look trivial that it’s not worth the cost of introducting ugly API
Trevor: will go off and come back with numbers

Ben: proposed making a new integration branch to merge all of the semver-minor features, plus a V8 upgrade, we can start pushing out nightly-style releases for playing around and punt the decision on how to do a major release down the road a little.
Group agreed, Rod agreed to help come up with a release mechanism for nightlies to couple with this.