Hi,
You're absolutely right. Look at it from both angles, customer side and service provider side. It is being said that you as a customer would set MED values to your outbound updates towards service provider in order to influence ingress traffic / on which path traffic enters customer AS; at the same time, the service provider could set MED values on inbound updates from the customer in order to influence egress traffic / on which path traffic enters customer AS. So, same BGP attribute, used to achieve same purpose, but applied in different directions and on different BGP peering side.
When you hear the statement, MED is supposed to be used in outbound updates to influence ingress path towards your AS and local preference is supposed to be used in inbound updates to influence egress traffic from your AS, both statements have been made from customer perspective, so that you as a customer decide the egress / ingress traffic paths, without any involvement from service provider side.
Considering the above statement, where we as the customer want to influence both egress and ingress traffic paths by doing BGP configurations on our side, without any involvement from the service provider, let's answer some questions:
1. Why is weight and local preference used to influence egress traffic from your AS, via inbound policies? Because weight is the first criteria in best path selection and local preference is the second criteria, so you want to use BGP attributes (weight is not a BGP attribute but for the sake of the explanation we consider it to be) that are at the top at the decision making process, to guarantee the outcome you're looking for.
2. Why is weight and local preference NOT used to influence ingress traffic to your AS, via outbound policies? Because weight is not a BGP attribute, is not attached to BGP updates, it's just a value locally significant to the router in its BGP RIB; because local preference, although it is a BGP attribute, so it gets attached to BGP updates, it is never attached to eBGP updates, only to iBGP, and as your customer to service provider BGP peering is of type eBGP, you can't use this attribute.
3. What are the next steps in BGP best path selection? Locally originated or not (and as we speak about manipulating egress or ingress traffic paths) this step is just ignored, as none of the BGP prefixes match this statement. AIGP, which is a big brother of MED, less known by engineers since it came later in the BGP development phases; it is called the big brother of MED because it's basically a simpler version of MED which gets applied higher in the BGP best path selection algorithm
4. Why is AS path and MED (mostly AS-path, MED is rarely used) used to influence ingress traffic towards your AS? Because these are the first BGP attributes, after AIGP (considering BGP best path selection algorithm) the can be used for this purpose, as AS-path is a well-known mandatory BGP attribute present in all BGP updates, while MED is an optional non-transitive BGP attribute which can be attached to both iBGP and eBGP updates.
Hope this make much more sense.
Thanks,
Cristian.