Hi Sheng, On 24-Aug-21 20:45, Sheng Jiang wrote: > Hi, Brian, > > I am fully support that there should be a mechanism to config GRASP in the > certain domain and also the flood methods. However, I have no clean the > definition of grasp-version and why.
I agree. This was intended as an example rather than a definite proposal. > Are we going to have multiple GRASP versions active in the same time? If (and only if) we added a version concept to GRASP, I suppose this could happen in a large network, with some nodes updated. > For my understanding, in an any giving time, we should have one fully version > of GRASP, then we may have many subsets of this full version of GRASP, which > may be combinations of functionalities/options of GRASP for different > scenarios. Or, perhaps, a new function (e.g. as suggested by draft-ietf-anima-grasp-distribution) would be supported in some networks but not in others. > Now, the question would be whether we, as ANIMA WG, should define these > subsets as standard with a registered number then easier to flood it in the > operation or leave them as individual functions/options but "GRASP manager" > can announce their availability on each function/option base. Exactly right. So the first question for the WG is: do we need such a capability? (At the moment, all we have to support subsets or unkown extensions is the M_INVALID message, so that one node can tell another "I didn't understand that".) Regards Brian > > Regards, > > Sheng > >> -----Original Message----- >> From: Anima <anima-boun...@ietf.org> On Behalf Of Brian E Carpenter >> Sent: Saturday, August 21, 2021 1:17 PM >> To: Anima WG <anima@ietf.org> >> Subject: [Anima] How GRASP could manage GRASP >> >> Hi, >> >> Following up on my previous message, here are some thoughts about how >> GRASP could manage itself. It will be a lame autonomic protocol if it can't >> manage >> itself ;-). >> >> One mechanism is that a "GRASP manager" ASA in an autonomic node associated >> with the NOC could send out configuration messages as GRASP floods. In all >> other autonomic nodes, a local "GRASP manager" ASA could receive these floods >> and act accordingly. >> >> For example, we could build this capability on top of what is already >> proposed in >> draft-eckert-anima-grasp-dnssd. Something like: >> >> objective-value /= { 1*elements } >> elements //= ( @rfcXXXX: { 1*relement } ) >> >> relement = ( relement-codepoint => relement-value ) >> relement-codepoint = uint >> relement-value = any >> >> ... >> >> relement-codepoint //= ( &(grasp-control:3) => grasp-config ) >> >> grasp-config = { >> ?( &(grasp-version:1) => 0..255), >> ?( &(max-flood:2) => 0..65536), >> ?( &(max-unicast:3) => 0..65536), >> } >> >> This allows us to specify a GRASP version for the domain (RFC8990 is version >> 1, of >> course) and maximum message sizes for flooding and unicast. This would be >> easily extensible for any other aspects of GRASP that we want to configure >> within a domain. >> >> (It's clear that we might also want to add an authentication/authorization >> mechanism to such messages, because they could be very dangerous if >> misused.) >> >> Is this idea worth following up? >> >> Regards >> Brian >> >> P.S. Thanks to Michael Richardson, Toerless Eckert and Carsten Bormann for >> some early discussion. >> >> _______________________________________________ >> Anima mailing list >> Anima@ietf.org >> https://www.ietf.org/mailman/listinfo/anima _______________________________________________ Anima mailing list Anima@ietf.org https://www.ietf.org/mailman/listinfo/anima