Comment 342 for bug 417757

Revision history for this message
In , hugh (hugh-redhat-bugs-1) wrote :

re comments #6 and #7: are these not bugs in your Juniper SSG-320M and load balancer? Should you not get them fixed?

I discussed this in https://bugzilla.redhat.com/show_bug.cgi?id=459756#c116

Ulrich: here's what I understand your fix to do: if the first AAAA query doesn't get answered, never ask for AAAA records again. But since UDP is unreliable, this seems actually wrong. Wrong for everyone, not just those with broken networks.

To avoid damage from these broken devices, glibc would have to use a distinct port for each query in flight. That would take some book-keeping in glibc, I would think (unless this case is the only way more than one query could be in flight).