Date: Mon, 19 Aug 2002 20:39:04 +0900
From: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
| > | this may be an implementation issue, but who should decide it?
| "who" means "userland, or kernel".
Oh, sorry, I misunderstood (which I guess was obvious).
For the kernel the decision is real simple I think. If it is picking
a source address, don't use a deprecated one if it can avoid it (if all
addresses are deprecated, or even all of the appropriate scope I think
the address selection draft says, then it has no choice).
Otherwise, the kernel should not even notice that an address is deprecated.
Assuming there's some reasonable way to make the information available to
userland, it should adopt the same policy. However, userland has done
extremely poorly on address management issues to date, so I am not
all that hopeful.
This doesn't make a big difference really though - preventing userland
from binding to a deprecated address might help very occasionally to
steer the application towards a preferred address (one would hope there
would be a better API for this than a syscall error...)
However, apps that bind to addresses that later become deprecated still
need to be dealt with. Those apps should be able to discover that the
address they thought was correct, no longer is, and rebind to the new one.
Of course, there aren't many such apps, most just use the address the
DNS says to use for remote addresses (or one passed from the peer) and
whatever address the kernel likes (or the one selected by the remote
partner) for the local address. Those just keep on keeping on...
kre
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
