> On Mon, Mar 4, 2013 at 1:30 PM, Dino Farinacci <[email protected]> wrote:
>> You make it scale up by starting to deploy something. That is where we are 
>> with LISP.
> 
> No, you're not.  You're thinking about maybe asking IANA to beg RIRs
> to hand out experimental address blocks.  RIR memberships are likely

Jeff, there are many paths to go forward. This ID is only a single path and if 
it doesn't pan out, it does not threaten LISP deployment.

We are just asking for a well-known prefix.

> to say no if it even gets that far.  If they happen to say yes, it
> will still create a very high barrier-to-entry for anyone to try to
> participate in what you claim you're deploying.
> 
> You're also saying that no one should be able to profit from EID
> registration so you reduce the potential pool of non-RIR-entities who
> might be willing to provide this service, and become stakeholders in
> LISP's success.

I am not saying that. I am saying I don't care how the business models are 
creative going forward. Let the market decide how this will go. And I certainly 
don't want to dictate or even suggest any models.

>> If there isn't a public mapping database system deployed, enterprises will 
>> do it themselves because it is easy enough to do. But we really don't want 
>> this to turn into many "private clouds" as we see with the various cloud 
>> based services being offered by industry. I think there will be private 
>> mapping database systems but I think there will also be public ones as well. 
>> The question is to try to avoid the complexities of a hybrid private/public. 
>> The same ones we see with cloud infrastructures right now.
> 
> I'm sure it hasn't escaped your notice that the DNS system has many
> registrars but only one delegation path for dot-com, etc.  Operation
> of mapping servers could be decoupled from the registration service,
> it could be a shared responsibility of several registration service
> entities, whatever.

Yes, it could.

> My point is that you haven't even thought about it.  All you seem to
> have thought is, oh the RIRs should do it for us, and by the way,
> let's make sure no one can profit from this.

We wanted to try using a process that is already in place. There is consensus 
forming that RIRs may be the wrong place. I'm fine with that.

> If two or more RIRs can provide registration services, then so can two
> or more ordinary businesses.  You know that, you've simply pretended
> it isn't true for the purposes of arguing against it in your post.
> That doesn't help anyone.

Umm, I'm confused. I made a suggestion. It could be way wrong. I wanted to 
generate discussion. We have accomplished that.

> You don't even have to sub-divide the EID block in the discussed
> manner to support multiple commercial registrars.

Agree.

>> Jeff, I see environments deploying overlays where all end-nodes are ONLY 
>> EIDs. So the negative map-cache entries don't even play there.
> 
> Yes, they'll just break xTRs.  See my previous postings on this topic.
> 
> As long as you pretend this is not a problem, you are obstructing any
> possible improvements.

The solution works, has limited scalability, and there are many ways to avoid 
or get around it. They come at costs, and we have to make tradeoffs.

Dino

> 
> -- 
> Jeff S Wheeler <[email protected]>
> Sr Network Operator  /  Innovative Network Concepts

_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

Reply via email to