Friday, 04/22

Attending: Zooko, Daira, Meejah, Brian, rkrishnan

• The new time (9am pacific) seems to work for everybody, but Zooko hadn't updated the calendar announcement that he maintains, so he and Daira hadn't heard about the new time.
• We landed a small patch from the magic-folders branch. Meejah continues to rearrange that branch, pulling out the pieces that are not specific to the magic-folders project, and submitting them as separate PRs (to land first). We're hoping to land the rest of the branch next week. One bug was filed (#2780) to use the pre-existing OneShotObserverList device instead of passing a Deferred into a class constructor: that cleanup should be pretty quick, but needs to happen before the branch will land.
• Zooko expressed concern about Tahoe's current behavior when it hasn't finished connecting to all storage servers: until we land the "servers of happiness" branch, a node will cheerfully upload all shares to a single server, or to itself (if that node is both server and client), if no other servers are available. This can happen if the upload is started very soon after node startup (before it's finished connecting to servers), or when the host computer is not yet on a network (think of a laptop with wifi turned off). He proposed that we actively discourage folks from using Tahoe on anything but a one-client-one-server configuration. That obviously didn't feel great, so we agreed to double down on landing shares-of-happiness (#1382). The branch itself is large and stale, and nobody really understands it anymore, so Zooko suggested the best approach might be to start by reading the docs, understanding the intent, then write a brand new branch (grabbing bits of code from the old one when possible), rather than trying to rebase/modify the existing branch.
• We closed several obsolete pull requests that had targetted intermediate branches, rather than master.
• Daira has rebased the cloud-backend branch to current master, and is working on making tests pass. She will submit a PR with the best branch to review, and we'll iterate on improving it (both to get the tests to pass, and to pull out any non-cloud-specific cleanups that can be landed first). My goal is to have a clear history for future readers: first turn the existing local-disk backend into a plugin, then second adding the cloud-backend provider plugins.

Tuesday, 04/26

Attendees: warner, meejah, dstainton

• reviewed, improved, and landed PR #269, which adds a hook to stall drop-upload/magic-folder uploading until a sufficient quorum of storage servers are connected. Discussed how to do this without using the recently-deprecated rref.notifyOnDisconnect()
• added a dependency on "mock" when running the test suite
• improve the comments in the initial tahoe.cfg file to explain that the "encoding parameters" are only used for newly-uploaded files, so you can safely change them whenever you want, without affecting your ability to download old files
• switched to https://appear.in/tahoe-lafs , which appears to use less CPU (perhaps it turns off the video codec when facemuted?), and caused meejah to be less robot-voiced than talky.io

The Tahoe-LAFS Weekly News is published once a week by The Tahoe-LAFS

The Tahoe-LAFS Weekly News is published once a week by The Tahoe-LAFS Software Foundation, President and Treasurer: Peter Secor . Scribes: Patrick "marlowe" McDonald , Zooko Wilcox-O'Hearn , Editor Emeritus: . Send your news stories to marlowe@antagonism.org - submission deadline: Monday night.