SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    RE: iFCP:Timeout and Encapsulation (was FC-BB exists, why invent something new?)



    Charles,
    
    > > Charles,
    > >
    > > Assurances are made with the fabric timeout but you do not offer a
    > > time-stamp to keep this promise.  TCP will try well beyond a FC fabric
    > > timeout.  A cable swap as example could exceed nominal delays.
    >
    > We are looking at a number of ways to measure whether or not the
    > underlying
    > network is performing within the parameters of such a performance
    > guarantee.
    > We're also considering a capability that allows the system user to set the
    > required timeouts and establish alarm thresholds and contingency policies.
    
    Should communication be interrupted and restored using various recovery
    techniques, this recovery may violate the time-out limitations important to
    FC.  In this case, heuristics of nominal communications would be of no use
    in determining if such an event occurred.  Other than placing a time-stamp
    within the header, I can not see how to police such a policy of squelching
    late frames.  Do you have a better idea?
    
    > > Would you see it possible to make a separate proposal to cover just
    > > encapsulation and another proposal to include iFCP specific
    > > link services?
    > > For those wishing just a tunnel, the encapsulation proposal,
    > > with a small
    > > (perhaps optional) iFCP specific field, could serve both purposes.
    > >
    >
    > We would need to carefully consider the consequences of such a proposal.
    > I'd be especially concerned about consequences down the road that might
    > impact the ability to make changes to both protocols over time.
    > Although I
    > can't speak for them of course, I suspect the tunneling folks may have a
    > similar view.
    
    A significant portion of encapsulation would satisfy both requirements.  I
    would recommend adding an option field in the same manner as TCP to include
    your additional requirements.  Do this in a way that allows both uses a
    level of extensibility as well as a means to negotiate a level of
    compatibility supported at each end.  This would ensure a larger community
    of products able to interchange and thereby improving your position within
    the market place.  You would still be allowed to add improvements that add
    value to your products and improve upon this scheme in the future.
    
    > <Material deleted>
    >
    > Charles
    
    Doug
    
    


Home

Last updated: Tue Sep 04 01:06:00 2001
6315 messages in chronological order