>>>>> On Tue, 18 Jul 2000 09:12:22 -0700,
>>>>> Steve Deering <[EMAIL PROTECTED]> said:
>> Fragmentation
>> One concern is regarding IPv6 fragmentation. There we use the source address
>followed by a unique ID generated at the source to define a fragmentation sequence.
>Now, consider the case where two link local addressed sources are sending fragments
>to a site-local address and happen to use the same ID's. I can not see a way in which
>the destination can re-assemble these two train of fragments properly.
>>
>> Flow Label
>> Same issue like fragment ID. Flow label ID and Src address again are used together
>to identify a flow. In the above scenario again, there could be issues in
>distinguishing across flows.
> Narsi,
> A packet is not permitted to leave the zone of uniqueness of its source
> address (see section 9 of the scoped architecture draft, second bullet),
> so your example problems ought not to arise.
Correct, but I recall another (or a related) issue triggered by the
original question. Consider the following topology:
nodeA-----------------nodeB------------------nodeC
fe80::1 fe80::2 fe80::2 fe80::1
This is valid if the link between A and B is different from the link
between B and C.
Then suppose that nodeA and nodeC are sending fragmented packets to
nodeB using a same fragment ID at the same time. A poor implementation
would be confused when reassembling the packets.
So, implmentations should care the zone of received fragments as well
as the source and destination addresses (and fragment ID) when
reassembling.
This may be a trivial requirement and needs not to be documented,
though.
JINMEI, Tatuya
Communication Platform Lab.
Corporate R&D Center, Toshiba Corp.
[EMAIL PROTECTED]
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
