[RRG] SHIM6: portability, ULAs, mobility etc.

[Robin Whittle <rw@firstpr.com.au>]

Thanks Marcelo for helping me understand SHIM6, in the thread "On
the Transitionability of LISP":

>>              SHIM6  Six/  Mobile  LISP-   LISP-   eFIT-  Ivip
>>                     One   IPv6    NERD    CONS    APT
>>
>> Address
>> portability                       Y       Y       Y      Y
>>
>> Multihoming   Y     Y             Y       Y       Y      Y
>>
>> Mobility                  Y                              Y*
>>
>> IPv4 too                          Y       Y       Y      Y
>>
>> No host                           Y       Y       Y      Y*
>> changes
>
> Just to point out that:
>
> Shim6 can support address portability, if you use portable
> identifiers, such as ULAs. No need to change anything in the shim6
> specs to support that.

My understanding of Shim6 is not ideal, but I don't know how this
could work.  How could host A send a packet to host B in the first
place if host B had a ULA address?


> Shim6 can be used as a RO mechanism for mobility, (it does not
> provides mobility anchor point capabilities, though)

I assume "RO" means Router Only.  I can't imagine exactly how this
would work.  Can you give an example?


> Shim6 can easily support v4 locators (but not v4 identifiers) (the
> NAT traversal capabilities would need to be worked out, but we are
> discussing about that)

Shim6 is therefore not of any help for IPv4 hosts trying to communicate.


> I don't know what kind of problem being solved is "No host
> changed" though... why didn't you include a line about "No router
> changes"? ;-)

I agree that no router changes is a definite benefit of SHIM6.
Likewise, along with Six/One, the lack of tunneling and the increase
in packet overhead, MTU and fragmentation limits etc. is a major
advantage over LISP/eFIT-APT/Ivip etc.

However, I am thinking about 99.9% of today's Internet users over
the next five to ten years.


> However, the main limitation of shim6 is about traffic
> engineering, that you seem to have missed. Shim6 do provides host
> enforced TE capabilities, but does not provides router enforced TE
> capabilitis, which seems to be a desired feature. (OTOH, Six/One
> seems to provide exactly that complementing Shim6)

I was looking at the bigger picture of helping the Internet as most
people use it today keep operating while a new router-based (a small
subset of routers) architecture is progressively introduced to limit
the 220,000 and growing load of prefixes on all the DFZ routers.

Shim6 and Six/One are not solutions to that problem.  They may well
be solutions to the long-term architectural problems of the Internet
*if* most users adopt IPv6.  I can't imagine how this would occur in
the next ten or fifteen years.

I know some people feel very positive about IPv6, but I find it hard
to be enthusiastic.

   The address is so long that all the headers are bloated.  Surely
   64 bits would have been enough, though this probably wouldn't
   support the innovative use of the many address bits which I think
   is central to Six/One.

   The whole thing is a separate system which can't simply be
   adopted as an upgrade, to replace IPv4.  If the network
   connectivity, protocols and interface to the operating
   system required a major switch, but could then fully
   support existing IPv4 communications as a subset of
   IPv6, then I think it would be better.  Instead, the
   host has to run two stacks and the hapless operating
   system and applications need to deal with them both,
   making tricky decisions.  I admit that I don't fully
   understand all the implications of dual stack on applications
   and on users, but from what little I know, it is messy and
   not at all seamless for users.

   The long addresses lead to much more work in the FIB
   of routers.  Likewise they lead to longer messages and bloated
   state everywhere.

99.9% of users don't know what IPv6 is and 95% wouldn't recognise
the term "IPv4".  They just want to use their computers and there is
no way of encouraging or forcing them to install new operating
system software, new applications or to get IPv6 connectivity by
some means.  They don't need to do this because every important
website and service is available via IPv4, and no-one would put an
important site or service only on IPv6 - unless it was only
important to a group of users known to have IPv6, such as perhaps 3G
cellphone users.

I think IPv4 will be the primary system for years to come.  I think
some LISP/eFIT-APT/Ivip system will be introduced and will enable
most of the address space to be closely populated with active users.
 I think this will keep IPv4 going for the next decade or two as far
as address space is concerned.  NATs are ubiquitous and will remain
so, unfortunately.

The pain of incremental change to something different such as IPv6
may forever be worse than the pain of incrementally investing in
more workarounds for IPv4's difficulties.  A few decades later, we
will look back and say "It would have been much easier to make
everyone switch to IPv6 or whatever in 2009 . . . ".


No doubt the IPv6 debate has been done to death and this is not the
place to rehash it.  I am just trying to explain why I don't think
it is such an elegant alternative, so people on this list understand
why I think most users will not be tempted to adopt IPv6 any time
soon.


  - Robin


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg