Re: Geofeeds are adversarial — was Re: Geofeeds are good

[Tom Beecher via NANOG <nanog () lists nanog org>]

> But the issue is that even with a million probe servers, IP geolocation
> can never be fully accurate in real time. How do we address the issue?


The large web companies/hyperscalers have been doing this for years. There
are functions in applications that send telemetry back from anyone who hits
their websites or opens their applications.  We can collect billions of
data points directly from our users, orders of magnitude more than you as a
3rd party could ever collect from just ping tests, and I can tell you that
none of us can reach 100% accuracy. Nobody ever will.


We chose to build a system that does not depend on ISP participation to
> produce accurate data.


You keep saying that your method 'produces accurate data' . But you provide
no evidence to support this claim.

Again, your company's business model is to sell access to a database of
information. All of the geolocation providers have their own method of
producing that data. However you (and all other geolocation providers)
shift the burden of CORRECTING that data to external parties , which costs
THEM time and money to do, just so you can flex about 'accuracy' and sell
more.

On Sun, Feb 15, 2026 at 2:03 AM Abdullah DevRel of IPinfo via NANOG <
nanog () lists nanog org> wrote:

> Hi,
>
> > Neither the end user nor the ISP has any way to know that IPinfo is at
> fault. Neither party knows that IPinfo even exists.
>
> You are right. IP geolocation should be invisible. Our approach
> fundamentally relies on the fact data is so good, we do not even exist. But
> the issue is that even with a million probe servers, IP geolocation can
> never be fully accurate in real time. How do we address the issue?
>
> > Your contractual relationship is only with the media distributor, and so
> is the end user's, and the ISP only had a contractual relationship with the
> end user.
>
> For our enterprise customers, we have established multi-channel escalation
> processes. I can describe what this looks like using Cloudflare as an
> example, since they are the largest internet company in the world:
>
> - Cloudflare's support and product teams have direct communication
> channels with our engineering and support teams.
> - Issues are escalated to us through automated processes.
> - Their community moderators and MVP users are aware of us and can flag
> issues directly.
>
> In the Cloudflare community, we see roughly one issue per month that
> involves our data, and most of those require explanation rather than
> correction.
>
> Example: https://community.cloudflare.com/search?q=ipinfo%20order%3Alatest
>
> Please understand that is Cloudflare and all their scale.
>
> OPNsense adopted our data as well. We made it clear to them that IP
> geolocation accuracy is our responsibility, not theirs. The issues I have
> seen in their forums have been cases requiring investigation and
> explanation, not systematic data failures:
> https://forum.opnsense.org/index.php?action=ipinfo
>
> Our accountability extends beyond paying customers to include users
> accessing our free data - IPinfo Lite (Country + ASN). We are actively
> reaching out to firewall companies to encourage them to adopt our free
> data, because end-users often experience frustration due to inaccurate IP
> geolocation data services. So, we are willing to provide our data for free
> in order to support their end-users.
>
> > You are saying end users must inform you, but they don't know you exist.
> You are saying ISPs must inform you, but they mostly don't know you exist.
>
> I am not saying either must inform us.
>
> For end users: our enterprise customers are instructed to maintain an
> escalation process. When an end user reports a location issue, the customer
> identifies whether it is an IP geolocation problem and forwards it to us
> through an automated channel. We investigate and resolve or explain within
> 24 hours. The end user does not need to know we exist for this to work.
>
> For ISPs: this is the reason we invested in active measurement rather than
> relying exclusively on geofeeds. Many ISPs do not publish geofeeds, do not
> maintain them, and do not know which IP geolocation providers are serving
> data about their prefixes. We chose to build a system that does not depend
> on ISP participation to produce accurate data. When ISPs do engage with us,
> it helps. But their participation is voluntary, not required.
>
> Again, we appreciate all the ISPs for their due diligence and effort to
> keep their geofeed.
>
> > And when an ISP does inform you, you push back on them, saying you trust
> the probe network more than the ISP.
>
> When our active measurement data conflicts with a geofeed or an ISP
> correction, we investigate. If the ISP's information is more accurate, we
> fall back to their geofeed and increase its priority score in our system to
> avoid the same issue recurring. This is not pushback — it is a verification
> step.
>
> > If you think media distributors must inform the end user to contact you
> first in the event of an IP geolocation problem, you must write that into
> your contracts with them.
>
> We do not require end users to contact us. The escalation path is: end
> user contacts the service provider, the service provider's support
> identifies an IP geolocation issue, and it is forwarded to us. This is part
> of our operational process with enterprise customers.
>
> Again, you are absolutely not required to talk to us. Nobody is required
> to talk to us. And the only path towards that on a systematic and macro
> level is active measurement.
>
> Our data is getting more granular and accurate. It is a continuous
> investment. We are only limited by how universal adoption of our data has
> been.
>
> ISPs, end users, network operators who are not our direct customers — we
> monitor public forums daily. This includes NANOG, Reddit (
> https://old.reddit.com/user/reincdr/), Hacker News, and customer
> community platforms for services that use our data. When someone has a
> complaint about IP geolocation, even if it is not specifically about us, we
> investigate. That is how we entered this thread. The original discussion
> was about BGP communities, not about us.
>
> > You say you are trying to distance yourselves from the rest of the
> industry, but you are only saying this with your words, not showing it with
> your actions.
>
> I understand that perspective. Here is what we are doing:
>
> - We maintain public presence in operator forums and respond to
> complaints, including complaints that are not directed at us.
> - We have a self-serve correction portal: https://ipinfo.io/corrections
> - We have a support channel: https://ipinfo.io/support
> - We are expanding ProbeNet to improve measurement accuracy so that fewer
> issues arise in the first place.
> - When our data is wrong and an ISP or end user flags it, we fall back to
> geofeed data and adjust the priority scoring to prevent recurrence.
>
> I acknowledge that forum monitoring is not the same as systematic routing
> of complaints. We are working on improving discoverability. If you have
> specific suggestions on how we can make it easier for affected parties to
> reach us, we are genuinely interested.
>
> — Abdullah | DevRel, IPinfo
> _______________________________________________
> NANOG mailing list
>
> https://lists.nanog.org/archives/list/nanog () lists nanog org/message/FFG63JXFCJRM3MKNFQB6OW245PFY2RUD/
_______________________________________________
NANOG mailing list 
https://lists.nanog.org/archives/list/nanog () lists nanog org/message/6DUXGFPUGWHIY4CT3WQEVWOPMQ3P3LMK/