We would like to use yield and generators in our code; it would help clarify a lot of asynchronous control flow.
This requires node > 0.10. We will probably let RESTBase be the pathfinder here, but shortly after it switches to node 4.x we'd like to follow suit.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Arlolra | T119228 Switch Parsoid to use node > 0.10 (most likely node 4.x) | |||
| Resolved | • mobrovac | T107762 Switch RESTBase to use Node.js 4.2 | |||
| Resolved | • GWicke | T91855 Offer io.js on Jessie | |||
| Resolved | None | T114399 Allow the build script to pick different nodejs/npm versions | |||
| Invalid | None | T123237 Provide production jessie image with node 4.2; use this for service-runner build command |
According to https://phabricator.wikimedia.org/T107762#1696377 node 4.2.x is the LTS release on the node side. Should we wait until this is present in an ubuntu release as well?
subbu notes that production is actually running debian (jessie), not ubuntu. But I'm thinking of downstream non-WMF users -- it would be best to wait until node > 0.10 is "commonly available" -- that is, available from both debian and ubuntu official repos -- before requiring it for Parsoid.
Practically speaking, we've had instructions for installing node from 3rd-party repos and/or using nave or nvm in [[Parsoid/Setup]] for a while. I'm just saying "general availability in popular distros" should be one of the factors we consider, not necessarily a determinative factor.
We don't have to require 4.x for Parsoid (by keeping our code working on node 0.10), but we could run the production cluster on it.
Change 319109 had a related patch set uploaded (by Arlolra):
T119228: Stop testing on node v0.10.x