Re: [Pce] 答复: Ramblings of two old dogs

["Adrian Farrel" <[e-mail]>]

Hello Oscar,

> I fully agree it is important to keep functional elements separated. That
> does not mean a vendor cannot sell a single box with several
> functionalities... 

sell != try_to_sell  :-)

[snip]

> ergo PCEP can be used to provide topology information to the
> PCE... Wait! Heresy!!! IGP does that! TED seems sacred and
> forbidden for PCEP... Syllogism failed... But it seems reporting
> LSP state information is OK....   

Hold on a moment.
Where does it say that the IGP is (MUST be) used for the collection of information in the TED?
In 4655 we have:
   The TED may be fed by Interior
   Gateway Protocol (IGP) extensions or potentially by other means.

Does that mean it makes sense to have each node in the network reporting TED information "northbound"? That depends. It doesn't make sense to me to have two protocols simultaneously doing the same thing in a network. But see BGP-LS.

[snip]

> Finally, Adrian, I would like to ask what would be the process of
> defining extensions of the PCEP protocol to be used by non-PCE
> boxes (and thus, out of current PCE charter).

That's two "finally"s in one email, but we'll let you off because you used "syllogism" in an IETF email, and that surely gains you some points.

The process would be easy...

If there is a WG for the work
   do the work there
   ensure PCE WG review for risks of breaking PCE
Elsif there is IETF support for the work
   do the work as AD sponsored
   ensure PCE WG review for risks of breaking PCE
Else the work to the ISE
   ensure PCE WG review for risks of breaking PCE
Endif

In general, re-use of protocols is a fine thing. Issues are only:
- breakage of existing uses (especially deployments)
- confusion in the market
- twisting or crowding that makes extensions for the
   primary use difficult.

Approaches used in the past for this type of thing have included:
- no distinction at all
- distinct ranges of TLV/message identifiers
- separation of protocol instances
- use of a new protocol number or port numbers even when the
   core protocol is the same

A