Hi Ketan,

A new revision of this document is now available 
https://datatracker.ietf.org/doc/html/draft-ietf-lisp-geo-18

Can you tell us if it addresses you concerns?

Thanks

Luigi

> On 17 Jun 2025, at 21:36, Dino Farinacci <[email protected]> wrote:
> 
> I have submitted -17, FYI.
> 
> Dino
> 
>> On Jun 17, 2025, at 5:55 AM, Ketan Talaulikar <[email protected]> wrote:
>> 
>> Hi Dino,
>> 
>> Please check inline below for follow-up with KT2.
>> 
>> 
>> On Fri, Jun 6, 2025 at 7:24 PM Dino Farinacci <[email protected]> wrote:
>> 
>>> KT> Looks like my comment was not clear. I am asking for a more detailed 
>>> specification of the mechanics that you are referring to in the use-cases. 
>>> Let us take an example:
>>> 
>>> This can aid in determining geographical distance when topological distance 
>>> is inaccurate or hidden.
>> 
>> You need geo distance when you may be testing for wireless signal range. 
>> Where topologically, at layer-2 or layer-3 is not relevant. We are also 
>> using geo-coordinates to locate things or to keep inventory of things.
>> 
>>> KT> How is this geographical distance exactly calculated? Any reference? If 
>>> not please specify. What is the unit of distance used? Does the document 
>>> propose to change the route calculation to shift from the topological 
>>> distance (coming from say underlying IGP?) to this geographical distance 
>>> now?
>> 
>> From a Geo-prefix, you build a sphere. If the geo-point is not inside of the 
>> sphere, you can determine something like "this device is not in San 
>> Francisco". So pick any point on the surface of the sphere and calculate the 
>> distance between the geo-point and any point on the sphere. To calculate 
>> distacne, see next paragraph.
>> 
>> When you want to know the distance between two points, there are geo 
>> libraries tht compute it based on roughly 100km going north and south 
>> (depending how close the points are to the equator) and same for longatude 
>> for east-west direction.
>> 
>> The unit distance is kilometers.
>> 
>> KT2> Let me take a step back. Getting Geo-Prefixes in EID and RLOC records 
>> looks to me as a protocol specification and not a use-case. Making map 
>> requests for these records and the map server returning the information 
>> (using physical distance) is also a protocol behavior specification - not a 
>> use-case. Therefore, I find all of this being under a "use-case" section as 
>> misleading and I am checking why this isn't covered as a specification - 
>> beyond just the encoding. Furthermore, "there are geo libraries that 
>> compute" does not seem to fully specify how this works - we'll need some 
>> references that point to how this calculation works or it can be specified 
>> inline. Are there any interoperability considerations here if different 
>> components of the LISP control plane were to interpret or compute these 
>> distances in a different way?
>> 
>>> 
>>>> ----------------------------------------------------------------------
>>>> COMMENT:
>>>> ----------------------------------------------------------------------
>>>> 
>>>> Please also find below some comments on this document:
>>>> 
>>>> Support of other DISCUSS positions:
>>>> 
>>>> 1) I support Deb's DISCUSS position related to the reference to 
>>>> OSPF/ISIS/BGP
>>>> individual drafts that carry similar extensions. Those documents have 
>>>> expired
>>>> several years ago. None of them were adopted by their respective WGs -
>>>> presumably due to lack of interest by the community? The case for changing
>>>> existing LISP extensions for geo-coordinates (also an experimental RFC8060
>>>> section 4.3) for the (sole?) reason of bringing consistency across routing
>>>> protocols does not look sound to me. Were there any 
>>>> implementations/deployments
>>>> of the encoding of RFC8060 that necessitate this change?
>>> 
>>> I hope Deb can accept Med's comments so I made changes to reflect those.
>>> 
>>> KT> That text change looks good to me. However, the justification of 
>>> changing the encoding that was in RFC8060 is not captured in this document 
>>> and that is what I am looking for. The point of consistency with other 
>>> protocols is moot (IMHO) since they don't exist today and those protocols 
>>> might as well use RFC8060 encoding (if/when someone really picks that up) 
>>> if consistency is all that we need. I hope you get my point.
>> 
>> It is beacuse we added more features to the encoding. Like cm granularity 
>> and radius to support Geo-Prefixes. ZSo the 8060 version is obsolete and 
>> hence we created a new code point in this document.
>> 
>> KT> Great. Could you please also add that as the reason for making the 
>> change over proposal in RFC 8060 (as opposed to the current consistency 
>> argument across routing protocols)?
>> 
>> Thanks,
>> Ketan
>> 
>> 
>> Dino
>> 
>> 
> 

_______________________________________________
lisp mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to