Re: [tor-dev] Tor and IP2Location LITE

On Mon, Aug 21, 2017 at 4:02 AM, Karsten Loesing <karsten@xxxxxxxxxxxxxx> wrote:

On 2017-08-16 21:19, Karsten Loesing wrote:
> On 2017-08-16 05:38, KL Liew wrote:
>> All,
>
> Hi Kim,
>
>> My name is Kim, the founder of IP2Location, a geolocation service
>> provider since 2002.
>>
>> It looks like Tor is looking to review other providers for GeoIP service
>> while I was reading one of a meeting minute for a meeting back in March
>> 2017.
>>
>> https://trac.torproject.org/projects/tor/wiki/org/meetings/2017Amsterdam/Notes/Metricsin5Years
>>
>> We are very interested in contributing to Tor and work on this matter.
>> Tor can host and integrate IP2Location LITE
>> (http://lite.ip2location.com) into their application. IP2Location has
>> programming libraries in most languages. We can also work with
>> developers if there is any technical issues.
>>
>> In term of accuracy, you can find the latest research paper published by
>> TUM. IP2Location has good accuracy as reported in Table V.
>>
>> Title : HLOC: Hints-Based Geolocation Leveraging Multiple
>> Measurement Frameworks
>> Authors : Quirin Scheitle, Oliver Gasser, Patrick Sattler, Georg
>> Carle from Technical University of Munich (TUM)
>> PDF Access : https://arxiv.org/pdf/1706.09331.pdf
>>
>> Let me know if there is any questions.
>
> Thanks for reaching out to us!
>
> It's indeed on our list to evaluate other geolocation databases and
> possibly switch over. I'll bring this topic up at tomorrow's metrics
> team meeting to discuss possible next steps for such an evaluation. I'll
> get back to you here to share the results.

So, we discussed this at our team meeting on Thursday and decided to
further evaluate switching to IP2Location.

That would be a non-trivial project, because we're using geolocation
data in at least two places: 1. shipped with the core Tor program and 2.
deployed on Tor Metrics services like Onionoo. And at least the former
requires close coordination with Tor's network team.

In any case we'll want to be sure whether this switch is the right move
before starting such a project. The paper is a good start, but we might
want to run more evaluations ourselves. For example, we could involve
relay operators by asking them which resolved location is closer to
reality. But even this evaluation requires writing some code, which puts
it on a long list of things we'd like to do.

All in all, we (Tor's metrics team) are considering it! But it'll be on
the order of weeks or maybe months before we can move this forward.

> One question, though, that just came to mind: Are there archives
> available for past IP2Location LITE databases, or do you provide just
> the latest version? Having archives, possibly even back to 2002, would
> be pretty useful for Tor Metrics. (I didn't look around as much on your
> homepage, so please apologize if this question is already answered there.)

You replied off-list:

> We do not have archive for the IP2Location LITE. We just started this free database a few years back.

Okay. Maybe we could do something with archive.org in that case. It's
not that we do have a complete history for MaxMind's files, except that
we could probably create our own history from Tor's Git repository which
contains files based on MaxMind's files.

All the best,
Karsten

>
>> - Kim
>
> All the best,
> Karsten
>