3 ms·

If you need more than country level data, these folks become very expensive very quickly. Many providers are orders of magnitude cheaper. I run one of them - h

If you need more than country level data, these folks become very expensive very quickly. Many providers are orders of magnitude cheaper. I run one of them - https://iplocate.io https://iplocate.io - but there are plenty of other high quality and affordable services.

tiffanyh 1y ago
Just curious, what do companies typically use as the source for such data?
- ignoramous 1y ago
  How to build a IP geolocation database from scratch, https://news.ycombinator.com/item?id=37507355 https://news.ycombinator.com/item?id=37507355 (2023).
Nyr 1y ago
They are not the cheapest, but most often will be one of the most accurate thanks to their latency triangulation.
- tallytarik 1y ago
  We and many others use the same techniques too. This is a concept introduced 20 years ago, and RIPE Atlas/IPmap has been a public implementation of this idea for the last decade or so. (However, the vast majority of IPs can't be geolocated in this way, and there are caveats to those that can be.) In any case, the difficulty with all providers in this space is how you prove accuracy at scale. If we assume some provider has some proprietary technique that nets 100% accuracy, that's great, but what do you compare it to? There is no ground truth data source - we are supposed to be that. Marketing plays a big role, and admittedly, these guys have much better marketing on this point :)
  - Nyr 1y ago
    I have no doubt that other providers use latency from probes as a data point. But IPinfo allows way more weight from this in their calculations, probably because they have developed their own reliable network, unlike most competitors. Your service, like many others, accepts as valid most intentionally fake geolocation data provided by networks. I am sure you know this anyway, so no need to mislead saying "we do the same".
    - tallytarik 1y ago
      This is an interesting point and in my view, there is no clear answer here. Here's an example: https://ipinfo.io/172.224.238.32 https://ipinfo.io/172.224.238.32 This IP is actually in Ontario, which you can easily verify with a ping measurement. But it is announced as being in Calgary by Apple's iCloud Relay geofeed (https://mask-api.icloud.com/egress-ip-ranges.csv https://mask-api.icloud.com/egress-ip-ranges.csv). Why? Because the intent of iCloud Relay is to obscure a user's IP address while still providing a roughly accurate location, specifically so that geolocation-based services still work as expected. For that to work, they need to provide 'fake data' in this geofeed so that they have pools of addresses covering thousands of cities around the world, AND they need geolocation providers to accept this. ipinfo accepts this, even though it's wrong. So do we. After all, geofeeds were supposed to provide a public geolocation database, the idea being that the network operator should be trusted to have the best information. We could provide the 'real' location, but if 9 providers say an IP address is in X and 1 provider says it's in Y, and Y is correct, you may just be frustrating end users of the network. But where is the line? I'm not sure, and it's hard to say who has the balance right here. We try to mitigate this by providing extra data like whether the address is a hosting or relay provider - for free, unlike others :) Some future addition could be to provide additional accuracy or source information, or even a 'reported' vs 'measured' location. We're working through this and hope to get to the right answer over time. Thanks for raising this. :)
    - reincoder 1y ago
      Thank you very much, Nyr. We really appreciate the trust you have placed in us. We are constantly expanding our probe network, and my colleagues are also working on stabilizing and improving the network. From 2025 and beyond, we will be putting a lot of effort into our R&D program. Our probe network provides the data that helps our data and research team to create better models. Currently, a good portion of the data is available to all with no compromise. When we do make mistakes, we hope our community of users will point them out to us. Each ASN or IP address mistake generates a new ticket, our CEO/Founder is tagged there, our data team investigates it, pushes fixes, and provides explanations.
- lpellegr 1y ago
  IPInfo appears to have good data, but their main advantage is strong marketing. They've effectively built a community of fanboys by publicly sharing techniques and methods that other major providers have been using for years or even decades.
  - reincoder 1y ago
    I would not say we have a super strong marketing effort, but we genuinely care about developers. That is how the company started, and that is how the company is operating after 12 years. We are super obsessed about any points of friction any IPinfo user has, but to be honest, that is just what any developer-first business should do. We are obsessed about developer sentiment and perception, but that is what the industry should be. Our users consist of the smartest people I know, and because we go above and beyond to be helpful to them, they do not point out mistakes we make, they actually try to help us. Consider our probe network server finding. We genuinely hit a wall after 750 servers. Then we reached out to our community of users, who found 150 more servers. There are even developers who will talk to their local hosting providers in their local language just to get us a server. Then writing code to integrate our data into different places. Our team is extremely small, so we cannot actively contribute engineering contributions to open source projects. Our users usually help us a lot in writing high-quality code for open source projects they already use. We are humbled because of the help our user share, and that is why we are obsessed with them and trying our best to go above and beyond to help them.

tiffanyh 1y ago

Just curious, what do companies typically use as the source for such data?

ignoramous 1y ago
How to build a IP geolocation database from scratch, https://news.ycombinator.com/item?id=37507355 https://news.ycombinator.com/item?id=37507355 (2023).

Nyr 1y ago

They are not the cheapest, but most often will be one of the most accurate thanks to their latency triangulation.

tallytarik 1y ago
We and many others use the same techniques too. This is a concept introduced 20 years ago, and RIPE Atlas/IPmap has been a public implementation of this idea for the last decade or so. (However, the vast majority of IPs can't be geolocated in this way, and there are caveats to those that can be.) In any case, the difficulty with all providers in this space is how you prove accuracy at scale. If we assume some provider has some proprietary technique that nets 100% accuracy, that's great, but what do you compare it to? There is no ground truth data source - we are supposed to be that. Marketing plays a big role, and admittedly, these guys have much better marketing on this point :)
- Nyr 1y ago
  I have no doubt that other providers use latency from probes as a data point. But IPinfo allows way more weight from this in their calculations, probably because they have developed their own reliable network, unlike most competitors. Your service, like many others, accepts as valid most intentionally fake geolocation data provided by networks. I am sure you know this anyway, so no need to mislead saying "we do the same".
  - tallytarik 1y ago
    This is an interesting point and in my view, there is no clear answer here. Here's an example: https://ipinfo.io/172.224.238.32 https://ipinfo.io/172.224.238.32 This IP is actually in Ontario, which you can easily verify with a ping measurement. But it is announced as being in Calgary by Apple's iCloud Relay geofeed (https://mask-api.icloud.com/egress-ip-ranges.csv https://mask-api.icloud.com/egress-ip-ranges.csv). Why? Because the intent of iCloud Relay is to obscure a user's IP address while still providing a roughly accurate location, specifically so that geolocation-based services still work as expected. For that to work, they need to provide 'fake data' in this geofeed so that they have pools of addresses covering thousands of cities around the world, AND they need geolocation providers to accept this. ipinfo accepts this, even though it's wrong. So do we. After all, geofeeds were supposed to provide a public geolocation database, the idea being that the network operator should be trusted to have the best information. We could provide the 'real' location, but if 9 providers say an IP address is in X and 1 provider says it's in Y, and Y is correct, you may just be frustrating end users of the network. But where is the line? I'm not sure, and it's hard to say who has the balance right here. We try to mitigate this by providing extra data like whether the address is a hosting or relay provider - for free, unlike others :) Some future addition could be to provide additional accuracy or source information, or even a 'reported' vs 'measured' location. We're working through this and hope to get to the right answer over time. Thanks for raising this. :)
  - reincoder 1y ago
    Thank you very much, Nyr. We really appreciate the trust you have placed in us. We are constantly expanding our probe network, and my colleagues are also working on stabilizing and improving the network. From 2025 and beyond, we will be putting a lot of effort into our R&D program. Our probe network provides the data that helps our data and research team to create better models. Currently, a good portion of the data is available to all with no compromise. When we do make mistakes, we hope our community of users will point them out to us. Each ASN or IP address mistake generates a new ticket, our CEO/Founder is tagged there, our data team investigates it, pushes fixes, and provides explanations.
lpellegr 1y ago
IPInfo appears to have good data, but their main advantage is strong marketing. They've effectively built a community of fanboys by publicly sharing techniques and methods that other major providers have been using for years or even decades.
- reincoder 1y ago
  I would not say we have a super strong marketing effort, but we genuinely care about developers. That is how the company started, and that is how the company is operating after 12 years. We are super obsessed about any points of friction any IPinfo user has, but to be honest, that is just what any developer-first business should do. We are obsessed about developer sentiment and perception, but that is what the industry should be. Our users consist of the smartest people I know, and because we go above and beyond to be helpful to them, they do not point out mistakes we make, they actually try to help us. Consider our probe network server finding. We genuinely hit a wall after 750 servers. Then we reached out to our community of users, who found 150 more servers. There are even developers who will talk to their local hosting providers in their local language just to get us a server. Then writing code to integrate our data into different places. Our team is extremely small, so we cannot actively contribute engineering contributions to open source projects. Our users usually help us a lot in writing high-quality code for open source projects they already use. We are humbled because of the help our user share, and that is why we are obsessed with them and trying our best to go above and beyond to help them.