Re: [[e-mail]: Re: BGP4 stuff: Local Preference Computation]

[Curtis Villamizar <[e-mail]>]

In message <c=US%a=_%p=SHL%l=SHL/CANADAW/[e-mail]>om>, "NITT
MANN Michael (MSMail)" writes:
> >
> >What you are saying doesn't make sense to me.  MED is non-transitive.
> >It is optional for EBGP.  It MAY be passed from EBGP into IBGP.  It is
> >NEVER passed from IBGP to EBGP.  It is perfectly legitimate to not
> >pass a MED at all in either IBGP or EBGP.
> >
> >Curtis
> 
> MED should be transitive and mandatory on all ebgp connections; doesn't look 
> so difficult to me.

Transitive is used in BGP to mean continuously passed from one AS to
another, usch as an ORIGIN or AGGREGATOR.  MED is not transitive.

> BTW, if I understand this right, and I always skipped over that part because 
> it is not relevant for a BGP state machine:
> 
> these 4 bits, what's the purpose at all? Are the property bits for 
> attributes in fact part of the transmitted protocol message (transitive, 
> etc....)?
> 
> Any real use transmitting protocol definition elements in run time PEs? 

The two bits are well-known/optional and well
transitive/non-transitive.  If you don't know what a "well known"
attribute means you have a protocol incompatibility.  If you don't
know what a new optional transitive attribute means you pass it on.
If it is non-transitive, you drop it.

The other two are extended length partial/complete and extended
length.  The latter is needed to figure out how long the attribute is.
If you didn't understand an attribute and passed it along you set the
partial bit to warn the next guy.

> Mike

Please read the whole draft before commenting further.

Thanks,

Curtis