IPv4/IPv6 fallback (happy eyeballs)

David Woodhouse dwmw2 at infradead.org
Mon Aug 14 01:49:20 PDT 2017


On Sun, 2017-08-13 at 16:40 -0500, Yang Yu wrote:
> It looks like if both A and AAAA record exists for a hostname,
> openconnect tries connecting over IPv6. It waits quite a long time
> before attempting to connect over IPv4. Any plan to implement happy
> eyeballs (RFC6555) like most browsers do?

There wasn't, but I suppose there is now... ;)

It wouldn't be fundamentally difficult — we already do a non-blocking
connect(), and poll for completion. So we already have the various
platform-specific difficulties resolved.

All we really need to do is expand the loop in cancellable_connect() so
that it has two outstanding connections at a time — one for IPv6, and
one for Legacv IP.

Realistically speaking I doubt I'm going to do this imminently. But
what I will try *really* hard to do this week is merge the other
outstanding work, including Daniel's GlobalProtect support, so I'm not
actually discouraging other people from contributing... 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 4938 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/openconnect-devel/attachments/20170814/6612d1aa/attachment.bin>


More information about the openconnect-devel mailing list