PMXBOT Log file Viewer

Help | Karma | Search:

#pypa-dev logs for Wednesday the 14th of May, 2014

(Back to #pypa-dev overview) (Back to channel listing) (Animate logs)
[02:19:09] <Ivo> dstufft: I'm wanting to customize your isolate() fixture to allow some tests to call it with funky unicode folder names
[02:19:27] <Ivo> or maybe I could just get it to use funky unicode folder names by default and see what fails...
[02:20:19] <dstufft> you could just do @pytest.mark.isolate_unicode
[02:20:55] <dstufft> and then do def isolate(request): if request.keywords.get("isolate", None):
[02:24:52] <dstufft> er
[02:24:59] <dstufft> request.keywords.get("isolate_unicode", None)
[02:25:46] <dstufft> the return value of that will have an .args and .kwargs, so you can do @pytest.mark.isolate_unicode("foobar") thing too
[02:42:50] <Ivo> awww
[02:42:53] <Ivo> its supposed to break
[04:18:45] <Ivo> dstufft: you know when pip {zip,unzip} can be removed?
[04:19:21] <Ivo> 1.7?
[04:19:43] <dstufft> https://github.com/pypa/pip/blob/develop/pip/commands/zip.py#L102
[04:19:45] <dstufft> yea
[04:19:49] <dstufft> they won't be in 1.7
[04:20:08] <dstufft> whatever number is in logger.deprecated is the first version without that thing
[04:33:36] <Ivo> dstufft: you ever hear of an acrobat account to do with python?
[04:33:52] <dstufft> I have no fucking clue what that ticket means
[04:33:56] <Ivo> haha
[04:34:20] <Ivo> I wonder if that's a spam bot?
[04:34:43] <Ivo> nfi what its business model is though
[04:36:24] <Ivo> dstufft: http://mynoise.net/NoiseMachines/catPurrNoiseGenerator.php and click on Animate!
[06:44:38] <Ivo> #$&*@#(U#@U$*@#U$ LOGS
[06:44:45] <Ivo> and their )$(@#*$ing encodings
[12:45:19] <jezdez> !logs
[12:45:19] <pmxbot> http://chat-logs.dcpython.org/channel/pypa-dev
[12:47:14] <jezdez> dstufft: that did the trick indeed, thanks for the quick fix!
[13:47:00] <dstufft> jezdez: awesome
[13:48:23] <dstufft> http://legacy.python.org/dev/peps/pep-0470/
[13:48:34] <dstufft> need to go over it again before I post it to distutils-sig
[13:54:14] <DanielHolth> if you ask nicely we can call you a cad without having to write and publish a whole document first
[14:01:46] <Alex_Gaynor> dstufft: don't use nginx. just use python
[14:03:12] <dstufft> Alex_Gaynor: huh?
[14:03:32] <Alex_Gaynor> dstufft: Why are you telling people to use a giant pile of C code, there are perfectly good webservers in python (which are easier to configure)
[14:04:16] <dstufft> Alex_Gaynor: because it's just an example
[14:04:35] <dstufft> also I know how to configure nginx to serve some static files correctly
[14:04:42] <dstufft> I don't know how to do the other things
[14:04:50] <Alex_Gaynor> twisted web --path=. :-)
[14:11:04] <dstufft> does that have autoindex
[14:11:40] <Alex_Gaynor> yes
[14:13:28] <dstufft> I don't think that command is right
[14:13:41] <dstufft> twistd web --path t just returns immediately
[14:13:45] <dstufft> and prints nothing
[14:13:50] <Alex_Gaynor> you got a daemon, add -n to not daemonize
[14:13:56] <Alex_Gaynor> sad default is sad
[14:14:22] <dstufft> ugh things that daemonize by default
[14:14:33] <Alex_Gaynor> ugh things that daemonize
[14:15:33] <dstufft> yea that too
[14:15:38] <dstufft> ok this works good enough
[14:15:41] <dstufft> I'll just do that I guess
[14:15:47] <dstufft> can I make it do TLS
[14:15:52] <dstufft> with some command line flag?
[14:16:01] <Alex_Gaynor> Yup. It even does /good/ TLS.
[14:39:27] <dstufft> Alex_Gaynor: http://hg.python.org/peps/rev/2e8893ae1431 there you go
[15:08:30] <dstufft> welp
[15:08:35] <dstufft> posted PEP470 to distutils-sig
[15:43:00] <pf_moore> dstufft: looks good - I hate to say this but directly cc-ing Stefan and MAL for comments might be worth it.
[15:43:43] <pf_moore> dstufft: although I don't see how it could be an issue for them. The people I see having problems are ones who host somewhere like Sourceforge and can't easily set up a viable index page.
[15:43:57] <pf_moore> dstufft: but they could just host on PyPI
[15:45:44] <dstufft> pf_moore: yea I might do that
[15:45:56] <dstufft> setting up an index page is pretty easy, it can be done with S3 or whatever too
[15:46:07] <dstufft> like, anywhere you can host static files
[15:46:48] <pf_moore> Sure. AFAIK, there's no requirement that a *custom* index can't have external links, is that right?
[15:47:34] <dstufft> if PyPI implements this, then I don't see any point to keeping around the code that tries to classify links
[15:47:38] <dstufft> so sure they could
[15:47:49] <dstufft> I doubt we'd keep around the code that does the indirect links
[15:47:55] <dstufft> so they'd have to be direct external links
[15:48:22] <pf_moore> Yeah, let's kill spidering completely, that'd be great
[15:48:40] <pf_moore> Hey, maybe PyPI could offer hosting for a project-specific "list of links" page that the project could use as their extra-index-url if they wanted. Unlikely anyone would ever use it, but it's there for people who have real issues.
[15:48:56] <pf_moore> More PR than anything else, but cheap & good PR, IMO
[15:49:07] <dstufft> well
[15:49:11] <dstufft> you essentially can do that already
[15:49:20] <dstufft> you can upload whatever you like on pythonhosted.org
[15:50:21] <pf_moore> Good point. I wondered if there was a way. Maybe add that as an example to the PEP then.
[15:50:46] <dstufft> though that might not work on pip with Python 2.7
[15:50:54] <dstufft> I think we might use SNI for pythonhosted.org's TLS
[15:51:01] <dstufft> and Python 2.7 doesn't support SNI
[15:51:15] <dstufft> unless you install pyopenssl ndg-httpsclient and pyasn1 anyways
[15:51:50] <dstufft> heh
[15:52:40] <dstufft> SNI is easy, normal TLS encrypts all the data, so there's no way for a webserver that hosts multiple sites to know what site you're trying to access.. so you can only have 1 SSL certificate per IP address:port
[15:53:04] <dstufft> SNI is just an extension that says, outside of the encryption stuff, also pass some data that says what host you're trying to access
[15:53:18] <pf_moore> Ah, I see
[15:53:20] <dstufft> that way servers can go "oh they want pythonhosted.org, I need to use a different certificate"
[15:53:47] <pf_moore> ... I think :-)
[15:53:56] <dstufft> it's basically the Host: header but for TLS
[15:54:22] <pf_moore> Gotcha
[15:54:34] <pf_moore> Would we require TLS for extra-index-urls, or just add a "careful, you're taking a risk" warning?
[15:55:07] <pf_moore> I can see the SNI issue being a problem for any shared-hosting scenario.
[15:55:38] <dstufft> I doubt we'll ever require it
[15:55:49] <pf_moore> That#'s what I thought - lots of people have said "I don't want to have to make my index secure and get a cert"
[15:55:54] <dstufft> we already have a warning that tells people "hey you're not using a safe transport"
[15:56:03] <pf_moore> I remember now
[15:56:12] <dstufft> eventually we'll have pacakge signing and using HTTP won't matter then either
[15:56:41] <dstufft> relying on TLS for our security is in my mind a stop gap until we get package signing
[15:56:52] <pf_moore> So pythonhosted.org/2.7 could just use http (it does offer unsecured http as well?)
[15:56:54] <dstufft> TLS is much much easier than package signing, so it was an obvious win
[15:57:06] <dstufft> yea pythonhosted offers HTTP and HTTPS
[15:57:39] <pf_moore> OK, so it's a viable approach for the odd cases. Cool.
[15:58:44] <pf_moore> Anyhow, must dash. Good job with the PEP, let's see how it goes down with the masses :-)
[15:58:51] <dstufft> see ya :)
[17:44:49] <Alex_Gaynor> dstufft: yay!
[19:49:38] <Ivo> Alex_Gaynor: python -m http.server
[19:54:22] <tomprince> That's probably not something you want to run in production, unlike twistd web.
[19:55:09] <Ivo> If I want to serve static files in production, there's no way I'd want to run a python web server to do that
[19:55:18] <Ivo> waste of resources