From: netmod <[email protected]> on behalf of Randy Presuhn <[email protected]> Sent: 10 March 2021 18:28 On 2021-03-10 12:43 AM, Juergen Schoenwaelder wrote: > Dear Qin, > > I believe this work repeats failures of the past but since the WG > agreed to entertain this, I will keep my mouth shut. I suggest you do > not spend your energy to convince the that this work is viable since > it is rather unlikely that I will change my mind.
<tp> Meanwhile the ITU-T has just liaised the IETF that it is starting work on intent-based management. I have not looked to see if the same words mean the same thing but guess that they do. Tom Petch I'm inclined to agree with Juergen. I participated in policy-based management work in ISO/ITU projects starting in the late 1980's, culminating in System Management: Management Domain and Management Policy Management Function (ISO/IEC 10164-19) as well as the later projects in the IETF and DMTF. Whether the problems are in the architecture, the specification, or implementation, there's an ample body of experience suggesting that this stuff is hard to get right, and to the extent that a proposal resembles any of the previous work, one would be well- advised to learn from the ways in which that previous work has crashed and burned, or in some cases never really left the ground. Have fun, and good luck. > YANG is for me _not_ a suitable policy language, it is at best a > language to carry policies written in a suitable policy language (and > I am not even sure about this). All attempts in the past to reach > agreement on a common usable standard policy language leading up to > interoperable implementations failed. The reasons are manifold but > strong (I think), standards-based interoperability at a generic policy > language abstraction layer is for me a myths. Complete agreement with Juergen. One *might* use YANG to describe the structure of policies, but it would be cumbersome for representing policy semantics. The current trend in YANG-land embracing ever- increasing deviation from "standardized" models means that standards- based interoperability is getting harder, rather than easier, when it comes to machine reasoning about the semantics of "related" models. > Please don't take this personal in any way. I just do not believe into > this work but I am happy to be proven wrong. Ditto, kinda. I still think policy / intent-based management is a laudable goal. But it's really hard, and there are many factors in both the IETF's technology toolkit as well as its standardization process that compound the difficulty of the problem itself. Randy _______________________________________________ netmod mailing list [email protected] https://www.ietf.org/mailman/listinfo/netmod _______________________________________________ netmod mailing list [email protected] https://www.ietf.org/mailman/listinfo/netmod
