We (Mixxx) have a user who reported an issue with hostname resolution. Here's the bug report: bugs.launchpad.net/mixxx/+bug/1391654 They say their hostname resolves to both ipv4 and ipv6 and they bound icecast to the ipv4 interfaces. (I'm not sure why they don't bind to both). Mixxx cannot connect to icecast presumably because libshout is resolving and connecting to the ipv6 address. We do nothing but call shout_set_host with the user-provided host string. They are requesting that libshout "fall back" on the ipv4 resolution of the hostname if the ipv6 version fails to connect. Two questions: 1) Is this intended behavior that isn't working? 2) Is this desired behavior to have within libshout? (i.e. a suitable feature request) If either is yes I'll go ahead and file a bug on the Trac. If not I'll keep this as a feature request for Mixxx. Best regards, RJ Ryan (Mixxx developer) -------------- next part -------------- An HTML attachment was scrubbed... URL: lists.xiph.org/pipermail/icecast-dev/attachments/20141117/4380df2a/attachment.htm
On 2014-11-17 7:55 AM, RJ Ryan wrote:> They are requesting that libshout "fall back" on the ipv4 resolution of > the hostname if the ipv6 version fails to connect.Binding only ipv4 on an dual-stack server is a deployment bug, but libshout should fall back when running on a dual-stack client, as they say. Please do file a bug. I believe RFC 6555 describes the current best practice for this: clients should probe over both protocols when dns returns both ipv6 and ipv4 addresses and use whichever to work around breakage like this (which can occur at intermediate routers as well as at servers). The http module in the opusfile library has an implementation of this if you want code to borrow. -r
Hm, I attempted to file a ticket but was rejected: Submission rejected as potential spam - BotScout says this is spam (Y|MULTI|IP|0|MAIL|0|NAME|2) - StopForumSpam says this is spam (username [40]) Maybe because I just registered? On Mon, Nov 17, 2014 at 12:59 PM, Ralph Giles <giles at thaumas.net> wrote:> On 2014-11-17 7:55 AM, RJ Ryan wrote: > > > They are requesting that libshout "fall back" on the ipv4 resolution of > > the hostname if the ipv6 version fails to connect. > > Binding only ipv4 on an dual-stack server is a deployment bug, but > libshout should fall back when running on a dual-stack client, as they > say. Please do file a bug. > > I believe RFC 6555 describes the current best practice for this: clients > should probe over both protocols when dns returns both ipv6 and ipv4 > addresses and use whichever to work around breakage like this (which can > occur at intermediate routers as well as at servers). > > The http module in the opusfile library has an implementation of this if > you want code to borrow. > > -r > _______________________________________________ > Icecast-dev mailing list > Icecast-dev at xiph.org > lists.xiph.org/mailman/listinfo/icecast-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: lists.xiph.org/pipermail/icecast-dev/attachments/20141118/1d435dc5/attachment.htm