Re: [PEPr] Proposal for Networking::GeoIP

Philippe Jausions wrote:

Have you performance tested the binary format vs. pure SQLLite (CVS import) database? Just curious here...

No, I haven't. All the tools provided by Maxmind (that I've looked at, anyway) work with the binary formats, so I haven't played with the SQL versions. Properly indexed, I assume that things could be pretty similar in terms of performance. SQLite does feel like the way to go for a PEAR package -- the portability is so nice (as illustrated by David Costa's MP3_ID class).

Also the GeoIP class does a lot more than ip-to-country mapping, so developing a driver-based system wouldn't make a lot of sense since the GeoIP driver would offer a very different API from the ip-to-country drivers.

Right. I have experience using MaxMind credit card fraud service, which however is web-service based.

Yeah, that's true -- there's a lot more beyond GeoIP too. And it is true that GeoIP is available as a web service. This class also doesn't deal with that possibility.

There is a Net_Geo package in PEAR which does only ip-to-country (afaik) using online services. It would make sense to me that other drivers that provide the same information could be added to this class. In fact Net_Geo could use Net_GeoIP for the GeoIP binary database format (Maxmind) rather than write its own parsing methods.

+1 on that.

Yeah, I don't know to what extend the Net_Geo package is maintained, but I think that it would be an ideal place to put a hot-swappable ip-to-country system. What I'm really trying to get is a pretty high-performance OO tool for designed to work with Maxmind's GeoIP stuff. I think that probably the target userbase for Net_GeoIP is a bit different from Net_Geo. Ideally, Maxmind could help with the maintenance of this API in the long run, as they periodically extend their db format, etc. Hans