ietf
[Top] [All Lists]

My Notes from Thursday Night Plenary - IETF 59

2004-03-04 06:34:54
Please contact me with any updates or corrections - thanks!

Spencer Dawkins

------------------------

Thursday Plenary - Leslie Daigle

Erik Nordmark - locator/identifier split

This concept is sticking it's head up from multiple holes, like a
gopher

- Want to start the entire community thinking about this concept

- One minute summary of multi-homing = sites connected to multiple
ISPs
    want to improve failure resilience, load balancing, better quality
connectivity
    today - addresses usually assigned by ISPs and aggregated toward
the default-free zone
    provider-independent addresses can't be aggregated by ISPs -
doesn't scale to one router per site in the world
    IPv6 could use IPv4 technique (but doesn't scale), multiple
addresses per site and per host (but has limitations)
    transport/application paths don't survive when there's a problem -
applications have to do recovery on their own
    don't know how to do address combinations - RFC 3484 is (only) a
start, because of ingress filtering

- Big questions
    separate identifiers and locators?
    current IP addresses would become locators - need an indirection
layer, but may not need a new name space
    one approach - ID->locator shim layer, but need a protocol to set
up this mapping
    works for most applications, but referrals and callbacks need help
    not sure where the shim layer goes - IP routing layer, IP endpoint
layer ...
    need a new namespace?
    FQDN as the key, sets of locators, ephemeral ID/purpose-built key
...
    stable long-term? survice ISP changes, etc.
    need a managed hierarchy? as IP addresses and domain names are
today? or self-allocated? hard to build a mapping function without
hierarchy
    don't know how to make these spaces scale without hierarchy -
could it be made to work in the self-managed case?
    how to "re-home" traffic? plenty of proposals to do this in multi6
    how to accommodate ingres filtering? if you can tell each ISP
about all other ISPs, this goes away, but won't happen for
consumer-class services
    need to select source/destination locators - when communication
starts, when communication fails. Is there SCTP experience we can
learn from?
    how to detect failures? depends based on application. transports
can give hints. routing system may have a role here. Or we could
heartbeat...
    new protocols should not make Internet less secure than it is
today, should not become the weakest link
    various threats exist (redirection, 3rd-party flooding - and this
includes amplifiers) don't depend on security pixie dust ("PKI").
    security threats that force hosts to do more work - vulnerable to
DoS attacks
    Mobile IP doesn't cleanly separate out multihoming - may be OK for
sites but not for hosts
    multi6 and HIP are working groups, HIP-related research group
forming in IRTF
    Erik asking for help in several areas

- Comments
    what about NEMO? NEMO is a consumer for multihoming, not an
inventor
    does HIP need hierarchy to scale? getting into details here
    thank you for reintroducing my proposal four years later...
    multihoming and mobility are mostly unrelated ... ???
    <renewal of multi6 working group meeting in plenary>
    structured and unstructured identity space - birthday problem is
real, even if it's 128 bits, for unstructured identity space
    separation between who and where is turning into a giant NAT - can
we remember peer-to-peer applications in use cases?
    this is an architected NAT when we rewrite headers
    tight presentation in a tough problem space
    multihoming and mobility are almost the same ... !!!
    presentation assumed new identifier space - please be skeptical
about this in your work - could have different identifiers at startup
and in the association
    may also use the same mechanisms for renumbering
    site multihoming, but maybe host multihoming is the interesting
problem
    trying to avoid assumption of a new identifier name space, but
it's really hard
    need help in understanding the requirements from people in the
room - is this damage control for an application, or a feature?
    can you sketch some NON-requirements? path selection based on path
quality, for example - is this really requirement? not an integral
part of the problem
    there is complex interaction between IP layer, transport layer,
application layer - solving the problem for TCP, but nothing beyond
that
    don't think application layer sees a different interface whether
this is provided in IP layer or transport layer
    can applications make use of locator information as well as
identifier information - remember site-local? this is a terrible idea
    I think of these solutions as a routing overlay, but this requires
end hosts to participate in the routing protocol - not a bad thing
    rewriting things scare me - going down a path that's not
sustainable
    mobile IP gives you one identifier, but MIPv6 is working on
bootstrapping, so you're not tied to one identifier
    rehoming protocol - there are at least four being talked about
today - can we work together for a better architecture?
    can't solve mobility and multihoming if this involves boiling the
ocean
    what if there is no solution to this problem? there are
engineering tradeoffs here
    mapping a referred identifier to a locator is really hard in the
general case - this may be a researchy question for a while
    there are a lot of proposals in multi6. Code would be good.
Reading code would be really good.

Leslie Daigle - Administrative Restructuring
    followon to presentation at IETF 58, this is a status report since
November
    draft-iab-advcomm-01.txt
    need a single focus for making operational choices that affect the
IETF
    draft-daigle-adminrest-00.txt is a high-level architecture
requiremnts document - specification to follow
    need IETF's own administrative function - need a full-time,
professional manager
    need better administrative pieces like WG tools, cross-area tools,
etc.
    not done directly by IAB or IESG(!) - not our core competency!
    please send comments directly to the authors
    plan to move forward to specific implementation proposal
    will continue to report status monthly
    have received zero comments on any of these documents - are you
reading them?

Comments
    are other people involved? mostly Leslie and Harald, with other
people who are affected
    how can we contribute if we want to help? stay tuned for the
actual project plan, and thank you for volunteering
    is the board structure really a requirement? you're overloaded
now - this is actually "appointed by" IAB and IESG chairs
    board diversity to insure lots of stuff
    wow - this is a lot more than just workflow management software!
how big a budget, how long a timeline, ....
    could we start small and see if we need all this?
    our concern is on how easy it is to propose a standard - if it's
not easy enough, people will leave IETF
    how to proceed with tool development? too early to know details
now

Harald - IETF Mission Statement
    draft-alvestrand-ietf-mission-00.txt
    "any organization that needs a mission statement is in serious
trouble"
    at least we'll shut down some of the mailing list discussion about
mission statements :-)
    goal of IETF is to make the Internet work
    Cardinal principles - open process, technical competence,
volunteer core, rough consensus and running code
    if the IETF can't help the Internet it will shut itself down, and
I'm proud of that
    please read this document and tell us what you think
    want to publish before the next IETF

Comment
    concern that our focus is on making the Internet work, not on
making IP networking work
    concern that ITU is doing MPLS work in competition with IETF - but
they are referring to 27 of our drafts in their current work
    if our standards don't work on the Internet, why would anyone
believe they would work anywhere else?
    we over-constrained, causing parallel efforts elsewhere
    PWE3 WG drafts taken in by SG15, without prior agreement, because
of frustration with us - please follow up on the list
    comments back only to the authors? didn't acknowledge 3-page
comment (whoops!) - we don't do host-to-host, only require Internet
traversal
    we got passed on preemption badly...
    comments can go to IETF list to keep Harald alert, etc.
    our quality is lower than other standards bodies - this will make
us disappear
    participants not standards professionals - we're doing evolution
here, not revolution
    IESG, IAB, WG chairs must spend huge amounts of time - you guys
really ARE standards professionals...
    Harald - "I'm an engineer on temporary assignment"
    we need people who can set their careers aside for a period of
time - and then return to their careers
    "a life, a job, and the IESG - pick any two"
    document headed in the wrong direction? sense of the room is
"no" - will continue

Harald the General Area Director - Changing the way the IETF Works
    inward-facing piece of the puzzle
    we have angry people - a symptom that something is wrong
    we are committed to improving
    ICAR - help with "late surprise" issues - security,
internationalization have been problematic lately
    in startup mode now
    NEWTRK - make the standards track make sense
    thinking for a while to make sure we're doing the right thing
    PROTO - help with IESG review process
    getting others involved
    MPOWR - smoothing the rough edges of the WG process
    still finding the next steps
    EDU team
    http://edu.ietf.org
    proposals we should "just do" - change rules on mailing list
management, etc.
    PROTO and ICAR changing the way IESG works - so step carefully
    we're starting to change things. we're trying to move quickly and
be careful. This is hard. Your help is needed.

Comments
    catching a surprise "late" - what is late? incompatible changes
after WGLC
    Working group draft status = old PS, no wonder there's confusion
outside this room
    abandon the label RFC?
    WGS presented in NEWTRK
    Area Charters to help people find (for instance) VoIP with all the
groups working in the space
    It's not just people who think RFCs are standards - RFC 821 is
still a full standard, and 2821 is still a proposed standard
    when people believe us about SMTP, they're still wrong!
    IETF is not good at cost-benefit analysis - AAA was so late that
the incumbent technology was extended - and it's still broken
    need more planning to accommodate this
    too much time and energy to send stuff back to Historic - should
happen automatically
    what we have in 2026 is reasonable - wish we followed it
    peer review in different WGs or even areas - we do this after last
call, too late
    training of WG chairs - everybody thinks delivering to the IESG is
the last step in the process, but the fun is only beginning
    EDU team is working on chair training, and PROTO team looking for
places to involve WG chairs
    20 percent of WG chairs come to training - and they get a free
lunch
    if there are too many documents to watch, stop watching all of
them - informational, individual submissions, etc.
    what working groups do I take this suggestion to?
    IESG will no longer take responsibility for technical review of
individual submissions - only end run detection - change will happen
in a month or so
    the July14 draft on process change will be pursued
    anyone can build code for anything, if they have enough money...
    how quickly will standardization process change? once no review
required to publish Informational RFC... is there a target? mostly
need a couple of months...
    IESG was designed to offload RFC Editor and IAB... we're where
were for the Cambridge tea party in 1992
    biggest problems are being addressed now, but no one is looking at
Internet Draft process - don't want to mandate xml2rfc, but could
automate submission...
    what documents the process, and the policies, for submitting an
RFC? I thought an Internet draft could be announced to a specific
mailing list, but no longer?
    announcements could be swallowed by spam filters
    secretariat staff has a document on how to process various types
of I-Ds - questions about formats and submission will be worked on in
the future
    individual drafts bounce from working group to working group for
multiple IETF meeting cycles
    it's hard to get attention for an individual draft - could ADs
help?
    in many cases when a draft bounces around, it's because the topic
crosses working group, or working group not able to take on new work
    can you help with existing working group items?
    just because a document isn't in a working group doesn't mean that
you can't be making progress outside a working group
    can we move to a more document format? please work this on a
mailing list. You will be flamed, but then you'll know...
    draft-klensin-process-july14-00.txt on process change? Update is
coming, watch for the update, feedback on solutions(_at_)alvestrand(_dot_)no




<Prev in Thread] Current Thread [Next in Thread>