If I can trust the User-Agent, I'm all set, and none of the rest
matters,
except the question as to whether there's a list of all possible
values
and their capabilities. But here are more details if you want to go
further:
> What are you seeing on the first request vs subsequent requests ?
Here are two lines from my apache logs, where I've X'd out some fields
out of paranoia. I'm running the N95NSTL java code.
[07/Mar/2008:14:02:00 -0500] "GET /jin/index.cgi?device=Nokia
%20N95Mods&tz=GMT
%2d05%3a00&widgetid=XXX&widgetname=&deviceid=XXXX&carrier=XXXXX&lng=XX&lat=XXX&screenwidth=240&widgettype=0&wquery=null&wname=HeyWhatsThat
HTTP/1.1" 200 5073 "-" "Nokia N95Mods/Sprint/Version=2.1/Profile/
MIDP-2.0 Configuration/CLDC-1.0 UNTRUSTED/1.0"
[07/Mar/2008:14:02:11 -0500] "GET /jin/pan.cgi?
id=XXXXXX&random=0&deviceid=XXXXXX&carrier=XXXXX&device=Nokia
+N95&lng=XXXXX&lat=XXXXX&screenwidth=240&widgetid=XXXX&widgettype=0&wquery=null&wname=HeyWhatsThat
HTTP/1.1" 200 8923 "-" "Nokia N95Mods/Sprint/Version=2.1/Profile/
MIDP-2.0 Configuration/CLDC-1.0 UNTRUSTED/1.0"
The point is that "device=Nokia N95Mods" in the first request, which
comes when
the user picks my widget off the widget list, but "device=Nokia N95"
in the
second request, which is when my index.cgi page calls pan.cgi.
Notice that device= matches the User-Agent field (the final field on
each line)
in the first request, but not in the second. I believe from earlier
experiments
and conversations that the value shown on the second line comes from
the consumer account associated with the device, independent of the
characteristics
of the actual device.
> Also, I dont follow this ... "Previous behavior, I think,
> was to always send a value based on the user account. "
I think that at some point in the past, both requests used to show
the same device= value, but I can't be sure.
MK