Re: Satellites running IP

From: Daniel Shell ([email protected])
Date: Wed Jun 19 2002 - 08:46:45 EDT

  • Next message: Daniel Shell: "Re: Satellites running IP"

    Eric

    So the answer is no.

    Sorry for the short reply but I am on vacation will get you a more detailed
    answer later

    At 11:58 AM 6/18/2002, Eric Travis wrote:
    >Daniel Shell wrote:
    > >
    > > Is SCPS an IETF RFC on a standards track yet?
    > >
    > > Dan Shell
    > > Network Architect
    > > CISCO Systems
    > > Gobal Defense and Space Group
    > > Wireless/Mobile Networking/ Satellite
    > > 216 643 2422
    >
    >Hi Dan,
    >
    >How are things going at Glenn?
    >
    >I'm confused by what you are asking but you've been
    >affiliated with NASA/Glenn and familiarity with SCPS
    >goes back long enough for me to remind you (sadly, not
    >for the first time):
    >
    > SCPS is NOT a protocol, but it is a set of
    > recommendations covering four protocol layers
    > (or 3.5 if you consider security a sublayer).
    >
    >Lots of things in the SCPS recommendations are
    >covered by IETF RFCs that are standards or standards
    >track - those things include such obscure things as
    >IP, TCP, UDP, ICMP, etc.
    >
    >You knew all that, however.
    >
    >If you are specifically asking whether the enhancements
    >that contribute to TCP Tranquility are covered by an IETF
    >RFC (Standard, Standards Track, Experimental, etc.) - then
    >the answer is:
    >
    > Nope - nobody every claimed they were, and
    > nobody in known space loses any sleep over it.
    >
    >But:
    >
    > The SCPS options have valid IANA assignments
    > (Options 20-23) and the protocol 106 covers
    > operation when using the SCPS compressed TCP
    > headers.
    >
    > As an aside:
    > I happen to pass lots of traffic onto the Internet
    > using TCP-Tranquility and I'm at least as nice as
    > every other TCP flow that I observe.
    >
    >Had NASA (or any other Space agency) expressed
    >a non niche interest (backed with the commitment
    >of necessary resources) in deploying any Internet
    >derived technologies over the space link and
    >beyond, the TCP modifications probably would have
    >been submitted as IDs with the desirement/goal of
    >having them someday make Experimental RFCs[*];
    >
    >[*] If I remember correctly, you folks at GRC
    > stated several times that you were going to
    > take elements of the SCPS recommendations
    > into IETF as part of the ACTS 118X (or Next?)
    > activities - maybe I should be asking you
    > your question - eh? :o)
    >
    >As it stands, the question needs to be answered:
    >
    >Why waste the IETF's time and meeting space on such
    >a insignificant problem space (no pun desired) that
    >even the mission folk at the space agencies don't
    >realm consider more than research?
    >
    >Does everything need to be covered by an IETF Standard?
    >
    >As I've said many time before (to you and others):
    >for a terrestrial environment, the elements of the
    >TCP-Tranquility feature set are experimental and tailored
    >primarily for the space environment (but work well in any
    >lossy tetherless network):
    >
    > Pure Rate Pacing is only useful for reserved paths
    >
    > Selective Negative Acknowledgements are best for
    > environments far more lossy than should be on any
    > end-to-end path going over the Internet backbone.
    >
    > The Vegas modifications seem to work well in general
    > but there still *may* be some nasty edge conditions
    > in networks that would not be typical to the space
    > environment.
    >
    > Any bandwidth estimation techniques are
    > likely to be burdened by the same concerns.
    >
    > Separating congestion signaling from loss:
    > We've never advocated running a default loss assumption
    > other than congestion for anything sharing an end-to-end
    > path with a public network (that may have different
    > assumptions and not the necessary machinery to explicitly
    > signal congestion)
    >
    > Explicit signaling of loss (trends)
    > I seem to remember some traffic polling for interest
    > on this topic (this mailing list or end-to-end);
    > This is somewhat of a localized issue and maybe not
    > something that needs an IETF standard.
    >
    > Link outages are a local problem and probably not something
    > that requires an IETF standard.
    >
    >There is more, need I continue down the list or have I made
    >the point?
    >
    >We've never been looking to push these concepts on a wired
    >infrastructure, but rather have always advocated using them
    >in networks which interface to a wired infrastructure.
    >
    >The most valuable thing that the space community (and anyone else)
    >gets from protocol standards are standard *interfaces* - which
    >include important things like header formats;
    >
    >Further, standardized state machines seem to track across diverse
    >environments, but the behaviors not covered by those state machines
    >may not.
    >
    >It is not clear to many of us that the protocol behaviors which are
    >best for operating over a diverse terrestrial network are necessarily
    >the best behaviors for operation over a resource constrained, time
    >bounded delivery infrastructure that is prevalent in most space
    >environments (and more than a few tetherless environments)
    >
    >You've probably been dealing in networking for long enough that you
    >probably remember the stupid/evil days of GOSIP - right?
    >
    >Please don't depress me and claim that everything used in a space
    >environment needs to be blessed as an IETF Standard (or Standards
    >track) RFC?
    >
    >Do you think the behavior of all the traffic that you are seeing on
    >your networks is strictly covered by (and conforming to) IETF Standards?
    >
    >Faith in guiding principles *can* lead to many wonderful things,
    >Dogma *always* leads to disaster, grief and collateral damage...
    >
    >Now - what was your question again?
    >
    >Eric

    Dan Shell
    Network Architect
    CISCO Systems
    Gobal Defense and Space Group
    Wireless/Mobile Networking/ Satellite
    216 643 2422



    This archive was generated by hypermail 2b29 : Wed Jun 19 2002 - 08:53:10 EDT