Re: [lisp] Fwd: Lisp Portable Edge Multipoint Sockets

[Sharon <[e-mail]>]

Maybe terminology will clarify this the best.

Typically EID means address of hosts which can move between network locations.

Instead we want EID to mean address of communication objects: mp2p queues and p2mp channels, which can move between hosts.

This portability supports functional distributed programming models and simplified on-path compute aware networking.

On-path awareness becomes important in the absence of centralized clouds which hide DNS changes from clients, and contain the unpredictable latency of co-located HTTP redirects. 

These geo-distributed compute conditions are relevant for the new AR/VR edge applications, and this specification simplifies the description of elastically scaling them using LISP.

--szb

Cell: +972.53.2470068

WhatsApp: [phone]

On Nov 21, 2022, at 20:33, Dino Farinacci <[e-mail]> wrote:

Yes, everything you can define the better. Even a "socket" since its an overloaded term. Plus you want to define terms based on how the document is going to use them. So defining "EID" in the context of your document would make the ideas more clear. Like an EID in this spec defines an object versus a host.

I don't think there is a standard format for describing the API. Just calls and input and output parameter descriptions you should include.

As for eBPF, just describe what filters you are need to use to associate packet flows with an EID. And that this is a local matter and the flows are NOT put in the mapping system like it could be for the Multi-Tuple draft.

I am not sure who you are referring to but I know a lot of people that use eBPF. You should google for a list name and propose they review the document and definitely invite them to IETF.

Dino

Begin forwarded message:

From: Sharon Barkai <[e-mail]>

Subject: Re: Lisp Portable Edge Multipoint Sockets

Date: November 20, 2022 at 11:30:45 PM PST

To: Dino Farinacci <[e-mail]>

Cc: Fabio Maino <[e-mail]>, Albert Cabellos <[e-mail]>, Jordi <[e-mail]>, Alberto Rodriguez-Natal <[e-mail]>

Thats all very true!

Could you say this also on the list so we know theres workgroup interest in developing this?

What about the EID Queue/Channel API ?
Do we have any good existing format we can adapt/adopt ?

How detailed should be the eBPF implementation description, can we engage the eBPF people that came to the last IETF? I think you know one of them .. they want to RFC eBPF itself.

--szb
Cell: +972.53.2470068
WhatsApp: [phone]

On Nov 20, 2022, at 22:51, Dino Farinacci <[e-mail]> wrote:

I think you need a lot more terms defined in your Definition of Terms section. Any reference to a new concept must be defined, even "edge compute IOT".

And you need more on the "how" its done. You have plenty on what it is.

Dino

On Nov 20, 2022, at 7:35 AM, Sharon Barkai <[e-mail]> wrote:


Could you have a look?
Can make writing other edge application specs simpler..

<draft-barkai-lisp-pems-00.txt>

_______________________________________________
lisp mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/lisp