0

[Below is testing I'm doing before using the ASP.NET C# API's to get screen resolution in pixels for any given mobile User-Agent in the world, using ScientiaMobile/WURFL]. Also note: I've included additional examples to support my claims, but they aren't necessary reading to understand/interpret the question, so please don't feel daunted, as they ('UPDATE' and beyond) are there for archival/appendix use.

Just to show you what I mean, if you go to this URL where you can enter a user-agent and have the screen resolution displayed/returned:

http://tools.scientiamobile.com/?user-agent-string=

Now, enter the following user-agent and click Submit:

Mozilla/5.0 (Linux; U; Android 2.3.5; ru-ru; Philips W632 Build/GRJ90) AppleWebKit/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1

Or, just click on this link below (instead) where I've embedded the user-agent into the URL:

http://tools.scientiamobile.com/?user-agent-string=Mozilla%2F5.0+%28Linux%3B+U%3B+Android+2.3.5%3B+ru-ru%3B+Philips+W632+Build%2FGRJ90%29+AppleWebKit%2F533.1+%28KHTML%2C+like+Gecko%29+Version%2F4.0+Mobile+Safari%2F533.1

Now, as you can see, it claims the screen resolution is 320x480, now, a quick search on google for that particular phone model (Philips W632) yields a totally different answer, that being a screen resolution of 480x800 (these are all in pixels) as visible here:

http://www.phonegg.com/phone/2614-Philips-W632

Now, if it was wrong on just one item, I'd understand, but I've put in random (but real) mobile user agents (tested about 5 now) and they have all come out to be wrong. So, I am wondering if there is something wrong in the way I am using the framework (right now not using API, but testing on the above test page before I commit).

Here are more results/examples if you'd like:

Phone: Nokia 308, WURFL claims 176Wx120H, here:

http://tools.scientiamobile.com/?user-agent-string=Nokia308%2F2.0+%2807.55%29+Profile%2FMIDP-2.1+Configuration%2FCLDC-1.1+Mozilla%2F5.0+%28Series40%3B+Nokia308%2F05.80%3B+Profile%2FMIDP-2.1+Configuration%2FCLDC-1.1%29+Gecko%2F20100401+S40OviBrowser%2F2.0.2.68.13.8+UNTRUSTED%2F1.0

Google claims 240Wx400H here: http://www.gsmarena.com/nokia_asha_308-5010.php

Final one I'll do...

Phone: Nokia 3250, WURFL claims 90Wx90H, here:

http://tools.scientiamobile.com/?user-agent-string=OneBrowser%2F3.1+%28Nokia3250%2F04.60%29

Google claims 176Wx208H here: http://www.phonegg.com/phone/199-Nokia-3250

Any ideas? Am I using the software wrong?

UPDATE: Additional examples, Reading Not Necessary:

Just noticed, it's reporting correctly for the Huawei H88 3G, but still incorrectly for many other smaller screen devices!

Mozilla/5.0 (compatible; MSIE 10.0; Windows Phone 8.0; Trident/6.0; IEMobile/10.0; ARM; Touch; Huawei; H883G; HuaweiH883G)

Check these out! More FAILS:

Phone: Verizon HTC Droid Incredible 2 ADR6350, WURFL claims: 320Wx480H, here:

http://tools.scientiamobile.com/?user-agent-string=TweetDeck+1.0.7.4+%28ADR6350%3B+us%3B+Android+2.3.4%29%0D%0A

Google claims 480Wx800H here: http://pdadb.net/index.php?m=specs&id=2819&c=verizon_htc_droid_incredible_2_adr6350

Phone: Nokia E71-1, WURFL claims: 172Wx120H, here:

http://tools.scientiamobile.com/?user-agent-string=Mozilla%2F5.0+%28Java%3B+U%3B+en-us%3B+nokiae71-1%29+UCBrowser8.3.1.161%2F70%2F352%2FUCWEB+Mobile+UNTRUSTED%2F1.0

Google claims 320Wx240H here: http://www.gsmarena.com/nokia_e71-2425.php

Phone: SonyEricsson ST25i, WURFL claims: 128Wx92H, here:

http://tools.scientiamobile.com/?user-agent-string=UCWEB%2F2.0+%28Linux%3B+U%3B+Adr+2.3.7%3B+en-US%3B+ST25i%29+U2%2F1.0.0+UCBrowser%2F8.2.0.242+U2%2F1.0.0+Mobile

Google claims 480Wx854H here: http://www.phonebunch.com/phone/sony_ericsson_st25i_kumquat-296/

Erx_VB.NExT.Coder
  • 4,838
  • 10
  • 56
  • 92

2 Answers2

1

[spoiler alert - I work for ScientiaMobile]

WURFL is focused on the detection of devices based on their web browser user agent strings. With the exception of the two that were correctly identified, the rest of the samples you've provided are from apps. We have an update coming soon that will improve the detection of UCWEB, since it is a web browser (and an app), but things like the Droid Incredible example (actually it's the TweetDeck app) TweetDeck 1.0.7.4 (ADR6350; us; Android 2.3.4) are changing all the time, and the value of detecting the screen size of the device that hosts an app is limited.

If you have a compelling use-case for the accurate detection of this traffic, we're happy to have a chat with you.

SteveK
  • 996
  • 1
  • 8
  • 11
  • Thank you for your response. But I have tried all of the above UA's (actual UA's of real surfers coming to real websites) by entering them into http://www.handsetdetection.com and they correctly identified all UA's including screen resolutions 100% correct. I also tried all above with 51degrees.mobi and they identified all correctly except for the TweetDeck app UA. How hard can it be when the device name is included in the TweetDeck UA's UA string, i.e.: ADR6350? – Erx_VB.NExT.Coder Aug 06 '13 at 13:15
  • I definitely see your point about the simplicity of pulling out the model name and matching that. This technique does indeed work, but it's very vulnerable to false positives (imaging "ADR6350 4G" or "ADR635"), and is generally slower, since you need to have complex regex to match all the possible model name formats and locations. Also, once you get into the non-Western market, there are many model name conflicts, so we need to be very careful about how we match UAs. So, while I do agree that these methods do have some advantages, our advantage is performance and very low false positives. – SteveK Sep 06 '13 at 17:11
1

Luca Passani, ScientiaMobile CTO here. As most know, ScientiaMobile is constantly on the task of monitoring the mobile web for the arrival of new devices and browsers, and fine-tuning WURFL to achieve great device detection. Of course, there is no limit to how far one can walk along the long tail of UA strings found in logs and eventually find something that is not recognized by WURFL. It seems to me this is the case here. But that's fine. We will happily take the challenge :)

A couple of the devices mentioned are effectively not in WURFL, but they will be very soon.

As far as UCWeb goes, there's a big update coming up (API 1.5) that greatly improves the detection of the increasingly popular web browser. While API 1.5 is still not out, commercial customers who have requested it, have been provided with a functional beta version who introduced the new feature for them.

Other UAs are from Apps. We are going to have some news on App detection too in the near future, but throwing app uas into WURFL indiscriminately is not going to be what most WURFL users want, believe me.

Finally, we would be more happy discussing issues like this on our developer forum (and this also guarantees faster turn around time from the ScientiaMobile team, so it's a deal that benefits all parts involved).

Thank you Luca Passani

Luca P.
  • 1,021
  • 8
  • 20
  • Thank you for your response, I wrote a comment back to SteveK, didn't want to repeat the same comment here so feel free to read that one above. Up voted you both. – Erx_VB.NExT.Coder Aug 06 '13 at 13:16