I'm sure it still confuses some people, but the more common it becomes, the more everyone gets used to it.
The benefit differs across markets and audience though, so I agree it shouldn't be a rule of thumb for good design. At least not yet.
For example, you and I know that you can geolocate my ISP to Nagoya with an API call that won't take a milisecond. How about, instead, doing some JQuery magic:
"Asking Google for your location..." (2 seconds)
"Searching for apartments around you..." (2 seconds)
"Picking best apartments..." (2 seconds)
$("precalculated-results").show();
I think one of the problems JangoSteve had with his implementation is that people really don't like to see forms already filled in on their website (we've gotten a lot of complaints about it at our work as well)
I really like the solution that was shown in the related article. All of his Geolocations have a regular text section that says something along the lines of:
"Ship something from _San Jose, CA_. (Not in San Jose? Click Here)"
Which would then (I assume) replace text with a fill-in form using JS.
That would keep the "Why is there information in my form!" support mails down, while at the same time providing the GEOfunctionality and an easy way to correct mistaken information (Porn sites always think I'm in Kasugai, which is about 50km from where I live grin)
And be sure to save their inputted area in a cookie, so they don't have to re-correct your geocache every time they see your site!
Or, indeed, in a different country with a different native language. I worked for one of those for a while, in a job that involved a fair bit of official web browsing (on-line reference materials for various things). Within days of the shift to routing everything through corporate IT, I wanted to kill whoever invented IP geolocation. It's a great idea that should almost never be used in practice.
Strangely - porn banners still show correct city in UK, not Italy... maybe PayPal should buy the geo DB from them instead?
Or, rather, web application developers who suck at implementing geolocation well can be annoying.
It is clear that no matter how smart you are as a developer, IP-only geolocation can't work reliably, simply because of the proxy issue.
And really, it's just balance -- maybe some people will still get confused. But maybe the benefit gained by the people who get their location guessed correctly outweighs the detriment of those who get a bad guess and get confused.
counter to the whole "Don't make me think" mantra
The problem is that confusing people forces them to think so you're actually making it harder for them to do what they need done. Honestly I have never seen a site where geolocation was useful. I guess if I were in a location where I didn't know the zip code, or street address, or street name it would be useful if I were looking for, e.g., a good restaurant. But in that situation, I wouldn't be using the internet, I'd just ask someone.
I actually implemented automatic IP geolocation on RateMyStudentRental.com, but it confused a lot of people because they weren't expecting it.
You have people trying to search for rental housing near their school while they're back home over break, and people searching for something across town by campus while they're at work, etc. Even the people who were trying to search for something nearby still expected to have to enter in the address, so they were completely thrown off, which is counter to the whole "Don't make me think" mantra. And don't even get me started on the people searching behind a corporate firewall, where the corporate headquarters (and IT dept) were halfway across the US.
In the end, I had to remove the automatic IP geolocation. Perhaps this is specific to my niche. Perhaps this feature is still ahead of the time (as the porn industry often is). Either way, that's what happens when you must hold your website changes accountable to actual metrics and results.