I am the assigned Gen-ART reviewer for this draft. For background on
Gen-ART, please see the FAQ at
<http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
Please resolve these comments along with any other Last Call comments
you may receive.
Document: draft-ietf-mpls-tp-te-mib-09.txt
Reviewer: Elwyn Davies
Review Date: 10 December 2014
IETF LC End Date: 8 December 2014
IESG Telechat date: (if known) -
Summary:
Not ready. I am not a MIB expert nor an MPLS(-TP) configuration
expert, so I have
assumed that the details of the MIB specification have been addressed by
Joan's MIB Doctor review and the configuration examples have been
checked by an
MPLS-TP expert. However, the introductory text needs significant work
to make it
easier for people who are not MPLS-TP experts to come to grips with the
document.
The major issue with this document is that it appears to go against the
current IETF view
that SNMP is not a satisfactory solution for configuring networks. The
document has
effectively reiterated the RFC 3812 situation from 10 years ago when the
IETF held its nose
and said that SNMP was the only game in town even if security sucked.
As I mention below, this was extensively discussed a year ago but I
couldn't determine whether this draft was given
a special dispensation even though consensus seemed to be that writeable
MIBs were
considered inappropriate. If this is the case, it will be sufficient to
adapt the end of s1.
AFAICS this draft has not had the 'grammar review' recommended by its AD
about a year
ago and there are a number of areas in the object descriptions where
language
improvements would be desirable in addition to notes given here. That is
it still needs
the grammar review.
There are some more minor issues, particularly as regards the
description of the types
of tunnel in s1, the discussion of the contents of the MIBs in s6 and
possibly in regard to
the interaction with the standard interface table.
Major issues:
Support for configuration:
The issue of SNMPset (write) support was discussed at length in the
context of
the MIB doctors' review of v-05 of this draft. The consensus at that
time appeared
to be that provision of write support in new MIBs should be strongly
discouraged.
I am unclear what the outcome of that discussion in the context of this
draft was,
but it is clear that the draft is very much written from the point of
using the MIB
as a configuration tool, especially in s9 (examples of usage). Whilst
this technique
reflects the equivalent section is RFC 3813, that document was written
over 10 years
ago and considerable effort has been devoted to providing alternative
and more secure
ways of configuring networks in the intervening time. The body of the
document has not
been altered to reflect the discussion that occurred.
"Lip service" is paid to the discouragement of the use of SNMPset by the
addition of
para 3 of s1:
It is understood that SNMP SET is not used for MPLS configuration these
days, however the read-write and read-create option is still specified
for some objects as a way to provide the information model.
It is not clear to me why providing read-write or read-create options
enhance the
MIB or are "a way to provide the information model".
I note that despite the extensive discussion of this topic by WG members
and MIB specialists
back in December 2013, this issue was not flagged up in the Shepherd's
write up.
If it has now been agreed that the configuration "flavour" is to be
accepted as WG consensus
in spite of being contrary to current practice, it appears to me that
the word "configuration"
should be removed from the last sentence of para 2 of s1 and the last
para of s1 rewritten to
explain why write capabilities are used in a more convincing manner.
Minor issues:
Abstract/s1: Needs to be clear in both places that is is intended for
MPLS-TP, this:
-- Abstract and s1: s/transport networks./MPLS networks conforming to
the MPLS Transport Profile (MPLS-TP)./
s1: What is a Non-MPLS-TP (transport) network?
s1: The explanation of what the MIB extensions in this document actually
do is not clear to the
general reader. I suggest replacing para 2 completely:
OLD:
The existing MPLS Traffic Engineering (TE) Management Information Base
(MIB) [RFC3812] and Generalized Multiprotocol Label Switching (GMPLS)
Traffic Engineering Management Information Base [RFC4802] do not support
the management of transport network requirements of Tunnel end points
with non-IP based identifiers and static bidirectional tunnels. This
document focuses on static bidirectional MIB modules that should be used
in conjunction with [RFC3812] and companion document [RFC3813] for MPLS
Transport Profile (MPLS-TP) path configuration and management.
NEW:
As described in the MPLS Traffic Engineering (TE) Management Information
Base (MIB) definition [RFC3812], MPLS traffic engineering is concerned
with the
creation and management of MPLS tunnels. This term is a shorthand for a
combination of one or one or explicitly routed LSPs (ERLSPs) linking an
ingress
and an egress LSR. Several types of point-to-point MPLS tunnels may be
constructed
between a pair of LSRs A and B:
- Unidirectional with a single LSP (say) from A to B.
- Associated bidirectional consisting of two separately routed ERLSPs,
one linking
A to B and the other linking B to A. Together the pair provide a single
logical
bidirectional transport path.
- Corouted bidirectional consisting of an associated bidirectional
tunnel but with
the second ERLSP from B to A following the reverse of the path of the ERLSP
from A to B, in terms of both nodes and links.
Tunnels may be either statically configured by management action or
dynamically created using
a LSP management protocol.
The existing MPLS TE MIB [RFC3812] and the Generalized Multiprotocol
Label Switching (GMPLS) Traffic Engineering Management Information Base
[RFC4802] address only a subset of the
combinations of statically and dynamically configured tunnel types,
catering for statically
configured unidirectional tunnels together with dynamically configured
unidirectional and
corouted bidirectional tunnels. They are alsorestricted to to end point
LSRs identified
by IP addresses.
The MPLS-TP TE MIB defined in this document extends the MIB defined in
[RFC3812] to cover
all six combinations (that is adding support for statically configured
associated and
corouted bidirectional plus dynamically configured associated
bidirectional tunnels). It also
extends support to end points that are identified other than with IP
addresses.
This support is provided by a suite of four MIB modules that are to be
used in conjunction
with the MIBS defined in [RFC3812] and the companion document [RFC3813]
for MPLS
Transport Profile (MPLS-TP) TE path [configuration and]* management.
END NEW
* Note: Using the term 'configuration' is inconsistent with the
discussion in para 3.
See Major Issues above
s5, bullet 1: s8 in RFC 3812 discusses the use of the Interface Table
with MPLS. It should be
made clear here:
1. Whether, if the MPLS-TP Tunnel is treated as an interface, there is
any difference in the usage
of the ifTable fields as compared with standard MPLS. In particular if a
different interface type
should be applicable.
2. Are there any issues if all or (more especially) a subset of the
tunnels are not treated as
interfaces
Row pointer usage: s9 of RFC 3812 considers Row Pointer usage - are
there any differences or
additional considerations for these MIBs?
s6: Sections 6.1 to 6.4 are not MIBs (as indicated by the title and
prologue) but MIB objects in
the MPLS-TE-EXT-STD-MIB MIB. Please reorganize s6 accordingly. Looking
back at RFC3812, the
structure of the document in sections 5 and 6 appears to be more helpful
to the new reader with a
summary of the four MIB modules in one section (expanded from ss6.5 -
6.7) followed by notes on
the key MIB objects in each MIB as necessary. It would also be helpful
to clarify where objects in
this MIB are expected to parallel those in the basic (RFC 3812) and note
that they are indexed
equivalently (notably in the case of mplsTunnelTable and
mplsTunnelExtTable).
Nits/editorial comments:
General: Consistent use of bi-directional or bidirectional: Please
choose one and use it throughout.
General: The qualifier "sparse" is used incorrectly in several places
(e.g., sparsely extended).
Presumably the intention is that the extensions are "slight" or
"limited" . In most cases the term
can be deleted: The degree of extension will ultimately be in the eyes
of the implementer!
(s5 - 2 cases, s6.4, s7 - 2 cases, s8, s9.1.6, s9.2.9, s9.3.6, s12 - 3
cases)
General: The figures need number and captions. The numbers should be
used to refer to the figures.
s4, para 1: s/unidirectional tunnel/unidirectional tunnels/
s4, para 1, last sentence: s/re-use/re-use and extend/
s4, para 2:
OLD:
retained in the same document, instead of a separate document.
NEW:
all defined in this document, rather than split over several documents.
s5, para 1: s/This document identifies/The MIBs in this document satisfy/
s5, para 2: s/The MIB module supports/The MIB modules, taken together,
support/
s5, para 2: s/static and signaled/statically configured and dynamically
signaled/
s5, bullet 4: s/sparsely extended/extended/
s9.1.6, s9.2.9, s9.3.6: s/a sparse augmentation/a limited augmentation/
s5, bullet 5: It would be helpful to explain how node identification
differs from
standard MPLS to understand why this mapping is needed. There appears to
be text in s6.1 that might help
but it is too late to make it easy to understand this section.
ss10-13: It is conventional to give the RFCs used as references in the
docuement s/RFC 3812/[RFC3812]/ etc.
s14: This is a near duplicate of the security considerations in RFC
3812. It might be easier to reference
that section and add a note that extra tables may provide further
sensitive information.
A brief discussion of whether the non-IP case makes any difference would
be useful.
s16: At the very least RFC 3811, RFC 3812 and RFC 3813 are normative.
RFC 6923 probably is.
RFC 6370 and RFC 4802 may also be.
_______________________________________________
Gen-art mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/gen-art
