b10-dhcp6 doesn't not bind on Scope:Global inet6 address.

[Angelo Failla]

Thanks Tomek,
That for the background reasoning, it makes sense.
-- 
Angelo Failla

Site Reliability Operations Dublin
Pallotron at fb.com




On 3/14/14, 12:34 PM, "Tomek Mrugalski" <tomasz at isc.org> wrote:

>On 14.03.2014, 13:15, Angelo Failla wrote:
>> On a side note I think you guys should make the v6 server binding by
>> default to ALL configured v6 addressesŠ
>> At least just to make it consistent with the way the v4 server works.
>> The decision is on you, just providing some user feedback :)
>It was a design decision influenced by couple facts.
>
>The first one is that it that binding a global unicast address is almost
>equal to enabling server unicast by default. What whould we do if
>clients start sending traffic to that unicast address? Should we drop
>it, even though we received the message (that's what RFC3315 tells us)?
>
>The other more fundamental reason is that some people do not want their
>server to listen on all possible interfaces. There are valid reasons
>where you want to listen only on specific interface or on specific
>address on that interface (e.g. you can have multiple global addresses
>and one of them is reserved for DHCP use, so you can filter the traffic
>more conveniently).
>
>Also think about more dynamic setups. What if you have a system that has
>temporary (like ppp) interfaces that occasionally go up? Should the Kea
>server bind to it immediately (as soon as it detects new interface)?
>Should it periodically detect if there are any new global addresses
>appearing on already configured interfaces? These are tricky questions
>without any clear best answers. So we decided to be conservative. If the
>server listens on something, administrator has to explicitly tell it to
>listen.
>
>It's a design choice, really. We can either not bind globals and provide
>a way for users to tell what they want to bind or bind everything and
>provide a way for users to tell what they don't want to bind. We chose
>the former, because it is more conservative.
>
>But of course that's not something carved in stone. It is possible to
>consider adding a flag, something like "bind-all-globals" or
>"bind-to-new-interfaces". But we don't have anything like this planned.
>
>Hope that helps,
>Tomek
>
>_______________________________________________
>bind10-dhcp mailing list
>bind10-dhcp at lists.isc.org
>https://urldefense.proofpoint.com/v1/url?u=https://lists.isc.org/mailman/l
>istinfo/bind10-dhcp&k=ZVNjlDMF0FElm4dQtryO4A%3D%3D%0A&r=ScT1tCv0IwhfW9O3zK
>6EHw%3D%3D%0A&m=wC8ZdkwJaxHJZqYlDTfdAWsOi%2B2w66a9z3U9ILAR5QA%3D%0A&s=0c9c
>33b2814d6062e7abf284d6eef2365cd73cf2e5eea94ed4ec6ec75652028c