>>I don't get the sense that we have consensus on this, because some
>>people seem to think that scoped addresses are appropriate for use by
>>general-purpose apps.
>>
>>for instance, there's really no way that an application can effectively use
>>a scoped address in a referral to another host, since the app has no idea
>>whether the host that uses the referral has access to the same scope as
>>the party providing the referral. name-to-address mapping is only one
>>instance of this problem.
another example of complication due to the scoped address.
(previous example was FTP)
X uses xhost(1) to control accesses to X server. For instance,
% xhost +10.1.1.1
would let X clients from 10.1.1.1 to access X server.
xhost(1) itself is a X client. therefore, it can be anywhere - it can
talk with X server over the net.
xhost(1) transmits the address (like 10.1.1.1 above) in binary form,
so 32bit for IPv4.
if we simply extend it to take 128bit IPv6 address, we cannot use
scoped IPv6 address, because of two reasons:
(1) scope identification will be stripped off when xhost(1) transmits
the address to X server (since the protocol does not pass scope
identification
(2) even if we extend the protocol to pass around scope ID -
if the view of the scope differs between xhost(1) and X server
(i.e. xhost(1) and X server are on different node), it is not
possible for xhost(1) to deduce scope ID on X server.
one possible solution would be to make xhost(1) transmit addresses
as a string, and let X server decode it. even so, the user of xhost(1)
has to know the view of the scope on the machine running X server.
itojun
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
