Re: Forwarding Packets With Link Local Destination Addresses

[Alexandre Petrescu <[e-mail]>]

Le 08/01/2021 à 06:52, 神明達哉 a écrit :
> At Thu, 7 Jan 2021 20:05:47 +0100, Alexandre Petrescu
> <[e-mail]> wrote:
> 
>>> Thanks! I think that the following text from RFC 4007 covers the 
>>> issue:
>>> 
>>> "   A node that receives a packet addressed to itself and
>>> containing a Routing Header with more than zero Segments Left
>>> (Section 4.4 of [3]) first checks the scope of the next address
>>> in the Routing Header.  If the scope of the next address is
>>> smaller than the scope of the original destination address, the
>>> node MUST discard the packet. "
>> 
>> Are scopes bigger one than another?
>> 
>> In multicast that may be true, but in unicast it's hard to say a
>> scope might be smaller or bigger than another scope.
>> 
>> In unicast, there might be only two scopes: LL and global.  It
>> might be that there is no inclusion from one into another, or it
>> might be that a size comparison might not be possible.
> 
> RFC 4007 explicitly defines the size relationship between scopes:
> 
> There is a size relationship among scopes:
> 
> o  For unicast scopes, link-local is a smaller scope than global.

I do not disagree with you, and I think many more people than just me
understand that ok.

This is about the semantics of words.

'Smaller' scope is not clear what it means.  These unicast addresses
dont have a number in them that is called 'scope' and that could be
compared.

Other RFC uses also the term 'broader' scope, instead of 'larger' scope.

Other RFC also tells that the IID itself (not the address) could be of a
certain scope, which we usually assume to be a characteristic of the
address not just the IID part.  Or says 'local scope' vs 'universal
scope', neglecting that the Globe and the Universe might be not the same
thing.  A Global scope on an IPv6-enabled Mars lander might have a
strange meaning.

Maybe the scope comparison in unicast could make sense when there were
were site-locals, but no more site-locals.

It's the notion of type that makes sense in comparing unicast addresses,
not that of scope.

And, when trying to identify whether a particular address is of a
particular type one absolutely needs the value and the number of bits to
compare.  For LLs one does not have the precise number of bits to
compare, because some RFC says that is 64 others say that is 10.  In
this situation it is impossible to precisely answer whether a particular
address is a link-local address or not.

Of course, we could say  that all addresses that start with fe80 are LLs
but then there are too counter-intuitive examples too that one could
easily build.


> o  For multicast scopes, scopes with lesser values in the "scop" 
> subfield of the multicast address (Section 2.7 of [1]) are smaller 
> than scopes with greater values, with interface-local being the 
> smallest and global being the largest.

Yes, for multicast addresses the notion of scope is different than for
unicast addresses.  In these mcast addresses there is a number that
represents scope and that could be compared and issue results like
'bigger than' or 'smaller than'.

Alex

> 
> The above cited text should be interpreted with this definition.
> 
> -- JINMEI, Tatuya
>