:zany_face: Sartre is shy -- who'd have thought....
", "time": "2025-03-20T05:02:31Z"}, {"author": "Tony Przygienda", "text": "those were all optimists, I wait for the unavoidable Dostoyevsky once we get little bit down the preso ;-)
", "time": "2025-03-20T05:03:14Z"}, {"author": "Tony Przygienda", "text": "what about \"we keep forwarding packets out of spite\" ?
", "time": "2025-03-20T05:03:27Z"}, {"author": "Tony Przygienda", "text": "hey, 800G is here as well ;-)
", "time": "2025-03-20T05:12:19Z"}, {"author": "Rajeev RK", "text": "If we are signalling anyways, why not signal the actual constraints themselves, rather than the origin computed path, so intermediate nodes can make constraint compliant routing decisions, without an origin path constraining them
", "time": "2025-03-20T05:13:14Z"}, {"author": "Tony Przygienda", "text": "yes, there are attemps to do that via sometimes atrocous things (Jeff H will disagree) like special communities
", "time": "2025-03-20T05:14:33Z"}, {"author": "Tony Przygienda", "text": "the problem is then that you end up sending everything everywhere which blows out the control plane
", "time": "2025-03-20T05:14:53Z"}, {"author": "Tony Przygienda", "text": "and ultimately you have to compute the constraint and put it on silicon which looks roughly like what signalling put there
", "time": "2025-03-20T05:15:09Z"}, {"author": "Tony Przygienda", "text": "and you will discover like SR that, duh, you don't know whether your path works end2end w/o BFD;ing every hop
", "time": "2025-03-20T05:15:44Z"}, {"author": "Tony Przygienda", "text": "no free lunch is the real theme of this preso ;-)
", "time": "2025-03-20T05:16:03Z"}, {"author": "Rajeev RK", "text": "and this still has the issue of visibility -- the source cannot ever truly determine all viable paths without visibility on availability of all intermediate links, which is unrealistic, IMHO
", "time": "2025-03-20T05:16:49Z"}, {"author": "Tony Przygienda", "text": "also, if you redistribute all constraints e'where how do you know they could be satisified, i.e. your path been setup. Thtt's why you have signalling ERO and so on crancking back on failures of setup
", "time": "2025-03-20T05:17:06Z"}, {"author": "Tony Przygienda", "text": "@Rajeev, funy enough RSVP-TE works and works well at sizes you wouldn't believe likely ;-) And we have stuff like flood reflection where the scale is unlimited for all practical purposes. Also, seamless MPLS has been floating boats for 20 years commercially just fine ;-)
", "time": "2025-03-20T05:18:06Z"}, {"author": "Tony Przygienda", "text": "@Rajeev, also, I would venture, the real TE game starts only once one starts to solve protection and bin packing if you are interested enough in the topic to look it up (or maybe you're already deeply familiar with such stuff, no offense meant).
", "time": "2025-03-20T05:24:48Z"}, {"author": "Rajeev RK", "text": "@tony I consider myself very much an amateur, so am definitely gonna read up further
", "time": "2025-03-20T05:26:43Z"}, {"author": "Tony Przygienda", "text": "and once at it add multicast to it ;-) RSVP-TE is not bad although most of the \"soft state\" original idea is gone (which was all the rage then ;-). Q.2931 is an old standard but a very, very good signalling protocol. As to engineering metrics, it's eitehr ISIS wide metric stuff used today or PNNI which was even richer.
", "time": "2025-03-20T05:28:09Z"}, {"author": "Benson Muite", "text": ":clap:
", "time": "2025-03-20T05:28:10Z"}, {"author": "Rajeev RK", "text": ":clap:
", "time": "2025-03-20T05:28:12Z"}, {"author": "Tony Przygienda", "text": "having said all that. tactical traffic engineering is a good synthesis in certain scenarios and very, very often just throwing BW at the problem is easier/cheaper than any other thing ;-)
", "time": "2025-03-20T05:28:51Z"}, {"author": "Tony Przygienda", "text": "Ruediger wants BIER then ;-) somebody brings him a bottle
", "time": "2025-03-20T05:31:08Z"}, {"author": "Boris Khasanov", "text": ":)
", "time": "2025-03-20T05:41:27Z"}]