Re: [Rift] Final pass on readability/normative ...

["Pascal Thubert (pthubert)" <[e-mail]>]

All good, Tony, many thanks.

From: Tony Przygienda <[e-mail]>
Sent: mercredi 29 janvier 2020 05:48
To: Pascal Thubert (pthubert) <[e-mail]>
Cc: [e-mail]
Subject: Re: [Rift] Final pass on readability/normative ...

Answers inline.

Also, to hono(u)r the English idiom, I belatedly replaced "leafs" everywhere to "leaves" as to not leave (as you know me, pun fully expected and intended) us exposed to grammar nit-picking ;-)

On Tue, Jan 7, 2020 at 7:57 AM Pascal Thubert (pthubert) <[e-mail]<mailto:[e-mail]>> wrote:
Hello Tony:

As promised:


      Superspine/Aggregation or Spine/Edge Levels:  Traditional names in
      5-stages folded Clos for Level 2, 1 and 0 respectively.  Level 0
      is often called leaf as well.  We normalize this language to talk
      about leafs, spines and top-of-fabric (ToF).

I believe we should reword this, maybe as below:

      Superspine vs. Aggregation or Spine vs. Edge or Leaf:  Traditional level
      names in  5-stages folded Clos for Level 2, 1 and 0 respectively.  We
      normalize this language to talk about top-of-fabric (ToF), top-of-pod (ToP)
      and leaves.

ack



   De-aggregation/Disaggregation:  Process in which a node decides to
      advertise certain prefixes it received in North TIEs to prevent
      black-holing and suboptimal routing upon link failures.

Suggestion:

   De-aggregation/Disaggregation:  Process in which a node decides to
      advertise more specific prefixes Southwards, either positively to
      attract the corresponding traffic, or negatively to repel it.
      Disaggregation is performed to prevent black-holing and suboptimal
      routing to the more specific prefixes.

ack



   Bandwidth Adjusted Distance (BAD):  This is an acronym for Bandwidth
      Adjusted Distance.  Each RIFT node calculates the amount of
      northbound bandwidth available towards a node compared to other
      nodes at the same level and modifies the default route distance
      accordingly to allow for the lower level to adjust their load
      balancing towards spines.

Removing the first sentence:

   Bandwidth Adjusted Distance (BAD):  Each RIFT node calculates the
      amount of northbound bandwidth available towards a node compared
      to other nodes at the same level and modifies the default route distance
      accordingly to allow for the lower level to adjust their load
      balancing towards spines.


ack



   North SPF (N-SPF):  A reachability calculation that is progressing
      northbound, as example SPF that is using South Node TIEs only.

I think it is confusing to call the northbound computation a SPF since it is a 1 hop calculation. Relates to the next point:



   We present here a detailed outline of a protocol optimized for
   Routing in Fat Trees (RIFT) that in most abstract terms has many
   properties of a modified link-state protocol
   [RFC2328<https://tools.ietf.org/html/rfc2328>][ISO10589-Second-Edition] when "pointing north" and distance
   vector [RFC4271<https://tools.ietf.org/html/rfc4271>] protocol when "pointing south".  While this is an
   unusual combination, it does quite naturally exhibit the desirable
   properties we seek.


I’d replace "pointing north" with computing southbound routes and reciprocally for "pointing south" with computing northbound routes.

done



PoD-ToP
->
ToP
(twice)

too tedious to find the places. better you do yourself


|H<
  ->
|o<

ack



LIEs SHOULD be sent with a TTL of 1
->
LIEs SHOULD be sent with an IPv4 Time to Live (TTL) / IPv6 Hop Limit (HL) of 1

(same addition suggested in 4.2.3.1 for LIE)

ack


Nodes in the spine are configured with "any" PoD which has the same
   value "undefined" PoD hence we will talk about "undefined/any" PoD.
   This information is propagated in the LIEs exchanged.




“Nodes in the spine” is unclear; do you mean ToF nodes? My understanding is that: ToF nodes MUST be configured with “undefined”, ToP MUST NOT be configured with “undefined”, and a leaf node MAY be configured with “undefined”, in which case the leaf node learns from its ToP parents. (If I’m correct) maybe we should say just that?

clarified to

"

Unless ZTP as described in Section 4.2.7 is used, each node is
provisioned with the level at which it is operating.  It CAN be also
provisioned with its PoD.  If any of those values is undefined, then
accordingly a default level and/or an "undefined" PoD are assumed.
This means that leaves do not need to be configured at all if initial
configuration values are all left at "undefined" value.  Nodes above
ToP MUST remain at "any" PoD value which has the same value as
"undefined" PoD.  This information is propagated in the LIEs
exchanged.
"



This will not affect the correctness of the protocol expect preventing detection of certain miscabling cases.
-> s/expect/except/
This will not affect the correctness of the protocol except preventing detection of certain miscabling cases.

ack



normally the top spines in a PoD
->
normally the ToP nodes

done



4.2.2.1.  LIE FSM

Maybe listing the states in after the first sentence would be useful for the reader? Also introduce HAL, HALS and HAT acronyms before using them?
Also, the terms “lie” and “cleanup” are found in lowercase, maybe change all relevant to uppercase

Yes, good suggestion on states, added that. moved some sections to make a clearer, more fluid LIE intro. massaged everything into one way of writing as one-way, two-way, three-way.

HALS/HAT is all definedi n the ZTP section where it belongs more logically. Someone reading LIE will get too overloaded with too many acronyms too early.



“More precisely, a spine node represents two different”
Should we remove “spine” from  the sentence? Or define it?
Same goes for the use of “Spine xxx”, should we call them ToP xxx?

nothing needs to be done, "Spine" is in the glossary and well defined.



4.2.3.  Topology Exchange (TIE Exchange)

Maybe indicate that TIEs are exchanged only in 3-ways?

yes, added in LIE.  A very mild possible loophole was present and I added.
"

A node SHOULD NOT send out any topology information elements if the
adjacancy is not in a "three-way" state.  No further tightening of
this rule is possible due to possible link buffering and re-ordering
of LIEs and TIEs/TIDEs/TIREs.

A node MUST drop any received TIEs/TIDEs/TIREs unless it is in three-
way state.
"

It was not a security loophole, normally the nonces would prevent from any flooding
attack under 3-way state but the addition helps to prevent "loose" implementations.





Suggestion to add an “instance nb” to the TIE so we can build multiple RIBs like in RPL over a same topology, e.g.; to serve different tenants.

unnecessary, all included. ;-)  4.3.9 already says

"

Multiplexing of LIEs can be achieved by
either choosing varying multicast addresses or ports
on the same address.
"

Multi-instance just runs on differnt multicast addresses and/or LIE ports and TIE reception port is advertised anyway in the LIE. Hence nothing more than UDP de-mux is needed to run multiple instances. Not a theory, I happen to have seen a nice multi-instance implementation already operating ;-)

Multiple RIBs is a rathole, that leads immediatly into VPN land and multkple RIBs/demux can be implemented in so many ways we'll never find a common language. Every routing stack (and they've been a couple) I worked on had a different architecture to deal with multkple RIBs (if they had one ;-)




More tomorrow, starting at 4.2.3.3.1.5

All the best

pascal



From: RIFT <[e-mail]<mailto:[e-mail]>> On Behalf Of Tony Przygienda
Sent: dimanche 8 décembre 2019 22:21
To: [e-mail]<mailto:[e-mail]>
Subject: [Rift] Final pass on readability/normative ...

Only IESG outstanding reviewer's comments on the spec are @ this point in time improving general readability & fine comb on normative captions (plus going to format -v3 with pictures is up to me)

I have some volunteers already that are starting on reading sections & polishing, I'm calling here for more ;-) If you'd like to review a section for readability & check normative language please contact me and I'll synchronize. Resolution is X.Y or X.Y.Z section with exception of small stuff like Examples maybe.

On offer honorable mention in the acknowledgements section ;-)

Idea would be to have a -10 (hopefully last one) with pictures & review incorporated before next IETF.

thanks

-- tony