[16:13:28] <jezdez> Ivo: it's a flash based solution, the js based one is currently still in development ;) we'd basically need to install a desktop app to participate
[16:13:42] <jezdez> it's called vidyo and the tech that hangout is based on
[16:14:34] <Ivo> the js webRTC put-it-on-any-webpage one is cooler
[16:21:15] <jezdez> I am for the open web after all
[16:21:17] <dstufft> jezdez: it should work just the same as PyPI itself does, if it doesn't that's a bug - though make sure you're using https://warehouse.python.org/pypi and not some other url
[16:21:39] <dstufft> Ivo: yes, there is a roughly 10 minute window where that would work
[16:21:41] <jezdez> Ivo: in theory talkzilla should be getting there, https://talkilla.mozillalabs.com/
[16:24:47] <dstufft> I don't like compromises where I feel like the end user is ultimately getting the short end of the stick, and it feels like MLs tend to be mostly about compromises that work best for the people doing the discussion, not for the end users
[16:25:40] <dstufft> Also I think half my value add is i'm the crazy guy that makes people more likely to agree to the reasonable thing :V
[16:39:37] <qwcode> dstufft, btw, no updates from me on the wheel cache PR. what time I had this weekend, I used on a few PPUG updates. if at any pt, you want to tackle it, don't feel like you have to wait on or use my PR, just ping in the PR so I know not to double time it
[16:54:32] <jaraco> Ivo, (in response to earlier) - The bot already responds to pings from the server. It probably just needs to exit or reconnect if it doesn't receive a ping within an expected amount of time.
[16:55:35] <Ivo> jaraco: i'm not aware of any standard time for a server to send a ping. might also be a bug in the socket logic not detecting a closed / broken socket?
[16:58:07] <jaraco> Ivo, could be. I've had reports from other users with similar problems.
[16:58:17] <jaraco> We don't experience the problem in our corporate environment.
[16:58:40] <Ivo> everything probably works fine if the network things are operating is a nice stable one :D
[16:58:48] <jaraco> Possibly because we have a lot of IRC activity, but more likely because the network is more reliable and less rigorous about tearing down idle connections.
[16:59:12] <jaraco> The base project is here: https://bitbucket.org/jaraco/irc
[16:59:40] <jaraco> With the primary socket logic here: https://bitbucket.org/jaraco/irc/src/1ba584f5642cbd4227eb80844b04240ca3a1eb13/irc/client.py?at=default#cl-245
[16:59:55] <Ivo> I've been meaning to write an irc lib for the new asyncio lib...
[17:18:51] <jaraco> Ivo, perhaps. It's not obvious to me how that would work, because the Manifold is allowed to be connected to multiple servers simultaneously.
[17:19:21] <jaraco> So something would need to be responsible for keeping track of the proper number of connections.
[17:21:04] <Ivo> jaraco: ive implemented logic before where if https://bitbucket.org/jaraco/irc/src/1ba584f5642cbd4227eb80844b04240ca3a1eb13/irc/client.py?at=default#cl-563 to L569 happen, without a flag saying the bot has intentionally disconnected, then just self.reconnect() in say 5 seconds
[17:23:08] <Ivo> public internet irc servers defintely aren't reliable to stay up consistently over days or weeks
[17:31:24] <Ivo> jaraco: the select call doesn't watch for any exceptions in the sockets either
[17:39:24] <jaraco> Ivo, good insights. Feel free to file a ticket or PR. Otherwise, I'll look into it when time permits.
[22:32:09] <Ivo> dstufft: TIL from the pypa-dev ML that --allow-unverifiable also implied --allow-external :D
[22:32:47] <dstufft> Ivo: yes, it didn't ship that way originally, but I added it in
[22:41:38] <Ivo> dstufft: if you remember me doing namespace packages initially for cryptography_vectors, i never had any problems there with editable installs & the namespace package re pip issue #3... i wish the comments had actually listed what the problem was, rather than 'regularly don't work well together'
[22:42:34] <dstufft> Ivo: Right now is th eopenstack summit so it's a bad time, however the openstack folks are running into this problem with their oslo libraries
[22:42:47] <dstufft> and I believe that they are planning to transition away from namespace packages because of it
[22:43:08] <dstufft> dhellmann in #pypa is probably a good person to talk to
[22:43:13] <Ivo> dstufft: maybe you could ask them to file a bug & enough info so that others could know what exactly it was?