[OpenWrt-Devel] Regarding bridge multicast-to-unicast patch

[Linus Lüssing]

On Thu, Sep 11, 2014 at 05:42:18PM +0200, Steven Barth wrote:
> Well it seems to be still somewhat used in embedded stuff, e.g. my
> 5-year old HP ethernet/802.11g printer/scanner/whatever still uses
> MLDv1 / IGMPv2 to do its MDNS / SSDP business.

Yeah, that's probably because IGMPv2/MLDv1 is way easier to implement
than MLDv2, so as long as IGMPv2/MLDv1 aren't obsoleted by the
IETF, especially budget vendors will probably keep implementing it
even today.

> 
> Another idea would maybe be to translate MLDv1 / IGMPv2 reports into
> appropriate MLDv2/IGMPv3 equivalents before sharing them on the
> bridge to prevent the suppression.

Hm, IGMPv2/MLDv1 to IGMPv3/MLDv2 should be relatively easy. But
I'm currently not sure whether a router/querier will make and take
some wrong assumptions and actions by thinking there is no
IGMPv2/MLDv1 listener. What if there is one MLDv2 listener which
included source specific state messages in its state change report?
The querier would respond with a "Multicast Address and Source
Specific Query" to learn whether there are listeners left for
this multicast and source addresses. Would the MLDv1 listener
interpret this query and process this message as a "Multicast Address
Specific Query" or would it ignore it?

What about the many MLDv1 queriers we have right now through the
bridge code? What if someone sneaks in an MLDv1 listener somewhere
in your bridged network over an ethernet cable? (You'd need
report translation both on the AP and on bridges and maybe query
translation, too)
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel