Re: Last Call: IP Version 6 Addressing Architecture to Proposed Standard

[Daniel Karrenberg <]

Authors

Robert Blokzijl 
RIPE Chair

Daniel Karrenberg
RIPE NCC General Manager


Background

RIPE is an organisation founded in 1989 wich aims to ensure the
technical and administrative coordination necessary for the operation of
the Internet in Europe [ripe-01]. 

The RIPE NCC performs activities for the benefit of the Internet service
providers (ISPs) in Europe and the surrounding areas; primarily
activities that the ISPs need to organise as a group, although they may
be competing with each other in other areas.  In particular the RIPE
NCC, as Regional Internet Registry, allocates and assigns IPv4 address
space in Europe and surrounding areas [ripe-162].  The RIPE NCC started
operations in 1992.  It currently allocates address space to more than
800 local Internet Registries almost all of which are operated by ISPs. 
The number of local IRs is expected to reach 1250 in 1998.  There are
currently no indications that the number of local IRs will stop growing.

The authors have contributed significantly to the development of the
distributed Internet registry system which is used for the allocation
and assignment of provider based IPv4 address space today.  As such they
have ample experience with the development of address space allocation
and assignment policies.  One important element of current policies is
that the size of address space allocations is determined by previous
justified use of address space.  A prerequisite for this policy is that
the size of allocations can start small and increase or decrease 
according to previous justified usage [ripe-159].


Argument

We believe <draft-ietf-ipngwg-unicast-aggr-02.txt> is critically flawed
because it standardises address aggregation boundaries without any
explicit technical justification.  In particular the length of the TLA
and NLA fields are proposed to be standardised as fixed at prarticular
values with no technical justification for either the fact that these
lengths need to be fixed nor for the particular values chosen.  The lack
of technical justification is significant because the standardisation of
TLA and NLA lengths directly influences many elements of Internet
operations including address space allocation policies.  In particular
the TLA being fixed at 13bit length makes only 8K TLAs available per FP.
Consequently Internet Registries will not be able to use proven
allocation policies but rather engage in regulatory practises.  The
rules proposed in <draft-ietf-ipngwg-tla-assignment-01.txt> are clear
evidence of this.  Broad acceptance of such rules and their implementation 
is extremely unlikely unless there is convincing technical reasoning 
behind the constraints that necessitate the rules. 

Because of this critical flaw we request that the IESG not advance
<draft-ietf-ipngwg-unicast-aggr-02.txt> and
<draft-ietf-ipngwg-addr-arch-v2-05.txt> to proposed standard yet.  We
suggest that the IESG refer these documents back to the authors and the
WG with the request to provide technical justification for the placement
of aggregation boundaries and to consider making these boundaries
variable where technically feasible. 


References

The referenced RIPE documents can be accessed at

http://www.ripe.net/docs/ripe-xxx.html		HTML
ftp://ftp.ripe.net/ripe/docs/ripe-xxx.txt	ASCII
ftp://ftp.ripe.net/ripe/docs/ripe-xxx.ps	PostScript