On 17-Nov-00 Erik Nordmark wrote:
>
>> [ylg]=> Yes, that's exactly wee need, a socket option like :
>> int on = 1;
>>
>> setsockopt(fd, IPPROTO_IPV6, IPV6_DONTFRAG, &on, sizeof(&on));
>> to disable fragmentation for UDP/Raw sockets (and of course on = 0 to
>> (re)enable fragmentation which would be the default).
>
> I'll add it to 2292bis.
>
[ylg]=> Nice, thanks.
>
>> As we said in a previous message, F. Dupont and me, a useful option would be
>> a new option to getsockopt() to get the outgoing MTU. For example, it helps
>> tracemtu to discover the right mtu to start (otherwise, we have to set
>> IPV6_DONTFRAG option, and try successive MTU (choosen in a list of possible
>> MTU, ordered decreasingly) until we don't get EMSGSIZE error (an other pb :
>> does exist such a list, i.e. the (right) list of all possible MTUs?).
>
> Presumably such an option needs to take a sockaddr_in6 as an argument
> to specify the destination. We can reuse the proposed ip6_mtuinfo structure
> for this.
>
> Do you care to know whether the kernel has path mtu information or
> is just returning the mtu of the outgoing interface?
>
[ylg]=> No, I don't care. I'm only interessed in outgoing interface mtu. I
think that your proposal to reuse the proposed ip6_mtuinfo structure is a good
idea.
Regards,
----------------------------------
Yves Legrandg�rard
E-Mail: [EMAIL PROTECTED]
Date: 17-Nov-00
Time: 17:49:40
----------------------------------
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
