In your previous mail you wrote:
> => I agree, it should be a library function (still in advanced API
> but not as a getsockopt()).
>
> [EMAIL PROTECTED]
>
> PS: on some systems getsockopt() is not supposed to have an argument
> (ie. optval points only to a buffer for the result) then be careful.
Good point. getsockopt only does a copyout from the kernel.
=> and your precedent mail proposed something which doesn't work...
Do you have a proposal for a name of a function which retrieves the
path MTU for a destination/path?
=> getpathmtu() ? But see the postscriptum.
Thanks
[EMAIL PROTECTED]
PS: I have changed my mind: Yves Legrandgerard convinced me that the whole
stuff should have similar interfaces then a getsockopt() is better but
should return only the MTU (32 bit unsigned integer) for connected sockets
(or largest compatible interface MTU for unconnected sockets). Same
for the ancillary data version (where the address is redundant).
PPS: NetBSD has something similar named IP_ERRORMTU which gets (by
a getsockopt()) the MTU to use at the last EMSGSIZE error.
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
