Log file Viewer

[08:57:04] <ssbarnea> i wonder if there is a way to prevent google from indexing old versions of pypi published packages.

[08:57:40] <ssbarnea> when I google a package name, almost for use one of tee top pypi links would be towards a very old release of a package

[08:58:03] <ssbarnea> i wish it could keep inside the index only the main page (w/o version)

[11:09:48] <ngoldbaum> there are webmaster tricks to fix that for sphinx docs (rel=canonical tags) i wonder if it makes sense for pypi itself to do something similar

[11:12:51] <PSFSlack> <uranusjr> `rel=canonical` is probably not a good idea since each page is dedicated to that release version, and the latest version is *not* the canonical source

[11:13:11] <PSFSlack> <uranusjr> But I’m sure there’s an attribute somewhere

[15:16:01] <ssbarnea> apparently there are some ipv6 issues with pypi.org where you endup failing to install packages, anyone aware of that? on ipv4 it works fine.

[15:38:26] <tos9> broadly yes

[15:38:29] <tos9> specifically perhaps not

[15:38:52] <tos9> ssbarnea: specifically e.g. https://github.com/pypa/pip/issues/5374

[15:39:31] <tos9> there's 1 or 2 others that mention ipv6 I think in the backlog

[16:06:40] <ssbarnea> ipv6 seems to be a recurring issue but there is something pip could do to mitigate: fallback between ipv6 and ipv4 on errors. current retry does not do that and fails if the default is broken.

[16:07:30] <ssbarnea> a downgrade to iov4 option would save these occasional outages caused by CDN

[16:07:44] <ssbarnea> at least some of them

[16:09:08] <ssbarnea> on opendev we temporary hacked the /etc/hosts to force ipv6 use to avoid it.

[23:44:30] <PSFSlack> <uranusjr> That’s something requests should do; pip does not even contain any DNS resolution code and is ignorant of whether the connection is made with IPv4 or v6