49 contemporary user agents which stress-test adaptive detection in DetectRight, WURFL and DeviceAtlas By Chris Abbott DetectRight Limited chris@detectright.com http://www.detectright.com 1 st June 2010
Abstract 49 recent user agents were extracted from a contemporary user traffic stream and submitted to three device detection engines: DetectRight (DR) (http://www.detectright.com), WURFL (http://wurfl.sourceforge.com) DeviceAtlas (DA) (http://deviceatlas.com) Composite scores showed: DeviceAtlas: 37% WURFL: 46% DetectRight: 96%. Detection performance scores were: DeviceAtlas: 27% WURFL: 44% DetectRight 100%. Performance on key datapoints was highly affected by recognition failures. Report compiled 1 st June 2010
Devices and composite scores in this document Page Section Device DR WURFL DA 8 Android HTC Buzz 5/5 1/5 2/5 9 Android HTC Desire 5/5 4/5 0/5 10 Android T-Mobile G1 5/5 3/5 3/5 11 Android Google Nexus One 5/5 4.5/5 0/5 12 Android Huawei U8230 5/5 5/5 3/5 14 WinMo Orange SPV M600 5/5 0/5 0/5 15 WinMo Generic Opera WinMo 3/5 1/5 0/5 16 WinMo HTC T7385 Touch Pro2 5/5 1/5 1/5 17 WinMo HTC P4500 TyTN 4.5/5 3/5 2/5 18 WinMo Samsung GT-i8000 Omnia 2 4/5 0/5 4/5 20 BlackBerry BlackBerry 9700 5/5 4/5 5/5 21 BlackBerry BlackBerry 8520 Curve 5/5 4/5 5/5 22 BlackBerry BlackBerry 8310 5/5 4/5 5/5 23 BlackBerry BlackBerry 9650 4/5 0/5 3/5 25 Apple Apple iphone 5/5 4/5 5/5 26 Apple Apple ipod Touch 5/5 5/5 4/5 28 Samsung Samsung GT-i7000 Galaxy 5/5 5/5 0/5 29 Samsung Samsung GT-B7610 5/5 1/5 0/5 30 Samsung Samsung GT-B5310 5/5 4/5 3/5 31 Samsung Samsung SGH-D407 5/5 4/5 2/5 33 Sony Ericsson Sony Ericsson Z520i 5/5 0/5 0/5 34 Sony Ericsson Sony Ericsson K790i 4/5 3/5 3.5/5 35 Sony Ericsson Sony Ericsson W810i 4/5 3/5 3/5 36 Sony Ericsson Sony Ericsson K800i 4.5/5 3/5 4.5/5 Page Section Device DR WURFL DA 38 Sony Ericsson K550i 5/5 2.5/5 0/5 40 Vodafone Samsung GT-B2100 5/5 4/5 5/5 41 Vodafone Vodafone V1240 4/5 4/5 0/5 42 Vodafone LG HB620T 5/5 2/5 4/5 43 Vodafone Vodafone 725 5/5 5/5 2/5 44 Vodafone Vodafone 810 5/5 0/5 0/5 46 Verizon ADR6200 (Eris) 5/5 3/5 2/5 47 Verizon ADR6300(Incredible) 5/5 4/5 1/5 48 Verizon UT Starcom XV6800 5/5 1/5 0/5 50 Sprint Touch Pro 2 (T7380) 5/5 1/5 1/5 51 Sprint PPC6900 5/5 1/5 3/5 52 Sprint PPC6800 5/5 1/5 5/5 53 Sprint PPC6850SP 5/5 1/5 0/5 54 Sprint MP6950SP/BM 5/5 0/5 0/5 56 O2 XDA Atom 5/5 0/5 2/5 57 O2 XDA Flint 5/5 3.5 0/5 58 O2 XDA Exec 5/5 2/5 1/5 59 O2 XDA II Mini 5/5 1/5 0/5 61 KDDI/AU Toshiba PLY 5/5 1.5/5 0/5 62 KDDI/AU Kyocera W65K 5/5 3.5/5 0/5 64 Nokia Nokia E72 4/5 1/5 3.5/5 65 Nokia 5320 XpressMusic 5/5 1/5 2/5 66 Nokia N95 5/5 1/5 0/5 67 Nokia N8-00 5/5 1/5 3.5/5 37 Sony Ericsson Xi1 5/5 0/5 0/5
Introduction Comparing device detection engines for accuracy has rarely been done, due to the difficulty of judging the results, and the difficulty of getting test data from outside the system. In addition, device databases tend to be interdependent, thus creating a self-referential incorrect consensus on the scale of the device detection problem space. WURFL and DeviceAtlas were chosen for this test because they are the most commonly used detection engines, and form the basis of numerous commercial and non-commercial services for large and small companies. I used a proprietary tool Detector Inspector to pick some real useragents from genuine contemporary web traffic to a rich media site. The user agent was usually newer than the data in the detection engine and thus a test of its performance on chaotic and new real-world data. Marks are given for each detection, with the mark starting from 5 and having one mark subtracted for each mistake. Generic detections with no redeeming features get zero. Where the marking differs from this, an explanatory comment is generally added. Comments are intended to consider why a particular engine failed or succeeded, and to illustrate particular points about the benefits and shortcomings of particular detection strategies. Comments are personal to Chris Abbott, and do not necessarily reflect the beliefs or opinions of DetectRight Limited DetectRight is a trademark of DetectRight Limited. DeviceAtlas is a trademark of mtld Top Level Domain Limited.
Component Scores Section Maximum DetectRight WURFL DeviceAtlas Detection Accuracy 48 100% 44% 27% Screen size 48 100% 43% 39% MIDP Version 49 100% 44%** 51% MP3 49 98% 59% 61% 3GPP Video 49 100% 51% 35% Browser Detection 48 98% 38% 0%* OS Version Detection 37 92% 38% 17% The figure is a 1/0 count of failures/successes in that area, expressed as a percentage. * DeviceAtlas does not give browser ID information ** WURFL does not have a field for MIDP 2.1 which impacts this score
Section Number of detections Composite Scores DetectRight WURFL DeviceAtlas Android 5 100% 70% 32% Windows Mobile 5 84% 20% 24% RIM BlackBerry 4 95% 60% 90% Apple iphone OS 2 100% 90% 90% Samsung 4 100% 70% 25% Sony Ericsson 6 92% 38% 37% Vodafone 5 96% 60% 37% Verizon 3 100% 53% 20% Sprint 5 100% 16% 36% O2 5 100% 32.5% 15% KDDI/AU 2 100% 50% 0% Nokia 5 95% 20% 45% All detections 49 96% 46% 37% Note: numeric score is based on recognition/detection performance and the accuracy of key datapoints. It s possible for a systemto fail the detection and score points, and for a system to recognise the device perfectly but still lose points on other items of data (for instance, the wrong OS version number, or not recognising that the device supports video). Since the useragents tested were all seen in the wild, it s reasonable to expect a device detection system to cope with them. DeviceAtlas does not attempt to give browser or version information, but has not been excessively marked down for that, though this data is importantin real life. WURFL data used was wurfl.xml dated 24/5/10, with PHP API 1.1 downloaded 17/5/2010. DeviceAtlas file was a licenced file dated 14/5/2010 with data from 08/05/10.
Android Devices Android device useragents are quite challenging for a detection engine. Android devices are proliferating, and devices are frequently upgraded. This means large numbers of unique or novel user agents hitting engines. Detecting OS version is pretty important for capabilities You can generate more Android detection tables like this at http://www.detectright.com/detector_inspector.html?request=android
HTC Buzz (Android 2.1) Mozilla/5.0 (Linux; U; Android 2.1-update1; en-us; HTC Buzz Build/ERE27) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 A very new HTC Device with same initial as another HTC Platform (Bravo). Which engine will generate a buzz? Detected As: HTC Buzz HTC Desire HTC Bravo DeviceAtlas over-identified again., WURFL picked the wrong similar string A8546ca761e5acd910d8c301ab520465 htc_desire_ver1 1840654 Mozilla/5.0 (Linux; U; Android 2.1- update1; en-us; HTC Buzz Build/ERE27) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 Mozilla/5.0 (Linux; U; Android 2.1-update1; xx-xx; HTC Desire Build/ERE27) AppleWebKit/525.10 +(KHTML, like Gecko) Version/3.0.4 Mobile Safari/523.12.2 Mozilla/5.0 (Linux; U; Android 2.1-update1; en-us; HTC B DA went up to HTC B before deciding on Bravo. The HTC Butterfly (if there was such a thing) would also generate this match. Screensize: 240x320 480x800 480x800 WURFL gets it wrong through detecting the wrong device. DA also makes the screen much bigger than it should be. MIDP version: 2.1 No 2 In this case DA misses a J2ME upgrade to 2.1 caused by Buzz being a newer device. WURFL doesn t give this a JVM. MP3 Support: Yes Yes Yes 3GPP Support: Yes Yes Yes Detected Browser: Mobile Safari 530.1 Android Webkit Not Available Detected OS: Android 2.1 Android 2.1 Android OS version missing for DA Performance Mark: 5/5 1/5 2/5 Not all HTC B-names are Bravo.
HTC Desire (Android 2.1 udpate 1) Mozilla/5.0 (Linux; U; Android 2.1-update1; en-de; HTC Desire Build/ERE27) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.1 HTC Desire is not a model as such, it s a hardware platform upon which other models are built. How will that work? Detected As: HTC Desire (Hardware Platform) HTC Desire Unknown Vendor Unknown Model 89c37301bd68f1d1c4645ce03bcf5152 htc_desire_ver1 DR and WURFL get this. DetectRight tries to differentiate between Hardware Platforms (reference designs upon which other organisations build) and Devices (things which are sold to the public). Mozilla/5.0 (Linux; U; Android 2.1- update1; en-de; HTC Desire Build/ERE27) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 Mozilla/5.0 (Linux; U; Android 2.1-update1; xx-xx; HTC Desire Build/ERE27) AppleWebKit/525.10 +(KHTML, like Gecko) Version/3.0.4 Mobile Safari/523.12.2 Mozilla/5.0 (Linux; U; Android 2.1-update1; en- DA gave up after the language! Screensize: 480x800 480x800 Not Available DA doesn t get this information. HTC Desire is a very good phone, so this is a bad miss. MIDP version: 2.0 No No You can sell apps to this device if you know it supports MIDP, but WURFL and DA don t pick this up. MP3 Support: Yes Yes No DR and WURFL get this right 3GPP Support: Yes Yes No Android supports 3GPP on all versions, if only to H.263. Detected Browser: Mobile Safari 530.1 Android Webkit Not Available It s debateable whether Android Webkit is a browser, since Mobile Safari is a brand, but WURFL is consistent. Detected OS: Android 2.1 Android 2.1 Not Available DR and WURFL share the spoils Performance Mark: 5/5 4/5 0/5 DA s detection is no object of desire.
T-Mobile G1 (Android 1.6) Mozilla/5.0 (Linux; U; Android 1.6; --; T-Mobile G1 Build/DMD64) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 What difference do dashes make? Detected As: T-Mobile G1 Android Dev Phone 1 T-Mobile G1 DeviceAtlas and DR got this right. WURFL chose the Android Dev Phone. Wrong, but relevant. ac8457fd37b7bb7336f25e5fd4aa3e59 android_dev_ver1_subandroid 16 Mozilla/5.0 (Linux; U; Android 1.6; --; T- Mobile G1 Build/DMD64) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 Mozilla/5.0 (Linux; U; Android 1.6; xx-xx; Android Dev Phone 1 Build/DRC83) AppleWebKit/525.10+ (KHTML, like Gecko) Version/3.0.4 Mobile Safari/523.12.2 1096144 Mozilla/5.0 (Linux; U; Android 1.6; - DeviceAtlas got this right: but for the wrong reason. Any Android string with OS 1.6 and a dash will be recognised as a T-Mobile G1. Given the flexibility and variety of Android strings, this is not a wise heuristic. Screensize: 320x480 320x480 320x480 Everyone gets this right: Android Dev Phone 1 was pretty close to G1 in specs. MIDP version: No No No UAProfiles do not mention MIDP support MP3 Support: Yes Yes Yes 3GPP Support: Yes Yes Yes Detected Browser: Mobile Safari 525.2 Android Webkit Not Available Detected OS: Android 1.6 Android 1.6 Android 1.0 A surprise here is DA s Android version number of 1.0, when the identifying string is version 1.6. Performance Mark: 5/5 3/5 3/5 DeviceAtlas s matching algorithm gets lucky, as does WURFL s.
Google Nexus 1 (Android 2.1 update 1) Mozilla/5.0 (Linux; U; Android 2.1-update1; sv-se; Nexus One Build/ERE36B) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 An updated Nexus One. How will the update affect the detection? Detected As: Google Nexus One Google Nexus One Unknown Vendor Unknown Model 257b78e130c69568e70e8ef3016b8031 google_nexusone_ver1_sub1 A complete miss for DA, an hit for WURFL. Mozilla/5.0 (Linux; U; Android 2.1-update1; sv-se; Nexus One Build/ERE36B) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 Mozilla/5.0 (Linux; U; Android 0.5; en-us) AppleWebKit/522+ (KHTML, like Gecko) Safari/419.3 Mozilla/5.0 (Linux; U; Android 2.1-update1; This is puzzling. DA at least knew by this point that it had an Android 2.1 device, but there s no OS level fallback data. Screensize: 480x800 800x480 Not Available Failure for DA. Not sure about that WURFL aspect ratio MIDP version: No No No MP3 Support: Yes Yes No Bad mistake here for DA, thanks to the detection miss. 3GPP Support: Yes Yes No Another bad miss for DA Detected Browser: Mobile Safari 530.1 Android Webkit Not Available Detected OS: Android 2.1 Android 2.1 Not Available DA s missed detection exacts a price Performance Mark: 5/5 4.5/5 (aspect ratio is odd) 0/5 Bad misses which would render an adaptive website very ugly indeed on Google s flagship device on DA
Huawei U8230 (Android 1.5) Mozilla/5.0 (Linux; U; Android 1.5; ru-ru; U8230 Build/CRB17) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 A much less common Android phone. Is this the droid* we re looking for? Detected As: HUAWEI U8230 Huawei U8230 HTC Magic A mismatch for HTC Magic, but WURFL spots it 65c49ea8e90b1cf20f9e86d931283b3b huawei_u8230_ver1 1426605 Mozilla/5.0 (Linux; U; Android 1.5; ru-ru; U8230 Build/CRB17) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 Mozilla/5.0 (Linux; U; Android 1.5; xx-xx; U8230 Build/CRB17) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 Mozilla/5.0 (Linux; U; Android 1.5; ru-r All Russians are Magic to DeviceAtlas. Screensize: 320x480 320x480 320x480 The HTC Magic and the Huawei are the same screensize: maybe the most common size for Android 1.5 devices. MIDP version: No No No MP3 Support: Yes Yes Yes 3GPP Support: Yes Yes Yes Detected Browser: Mobile Safari 525.2 Android Webkit Not Available Detected OS: Android 1.5 Android 1.5 Android Despite including 1.5 as part of the detection string, DA misses the version again. Performance Mark: 5/5 5/5 3/5 DeviceAtlas misses the device, but gets the core features right. * Droid is a registered trademark of LucasFilm
Windows Mobile Devices The main problem with Windows Mobile Devices used to be that there were no identifying UA features. These days, ID information can vary greatly from device to device, and there are multiple different browsers the device can run. Any engine which is not actively scanning the tokens of a useragent will generate many misses on novel/new useragents. You can generate more Windows Mobile detection tables like this at http://www.detectright.com/detector_inspector.html?request=windows%20mobile
Orange SPV M600 (WinMo 5.0) Mozilla (Windows Mobile; PPC; 240x320; M600; OpVer 13.21.15.104; Opera Mobi/35166; U; en) Presto/2.2.1 Opera/9.7 An identifiable Windows Mobile Device running Opera Mobile: how will everyone cope? Detected As: Orange SPV M600 Opera Mobi 4 Unknown Vendor Unknown Model Both WURFL and DR have a problem with this Windows Mobile running Opera 3b87c919dc4d1a7d191ebd2c4d7887 90 opera_mobi_ver4 WURFL detects this as an XHTML generic device, which for markup would be OK. Mozilla (Windows Mobile; PPC; 240x320; M600; OpVer 13.21.15.104; Opera Mobi/35166; U; en) Presto/2.2.1 Opera/9.7 Opera/9.7 (Windows Mobile; PPC; Opera Mobi/35166; U; en) Presto/2.2.1 Mozilla ( WURFL fixates on the browser Screensize: 240x320 176x160 Not Available DR gets this MIDP version: 2.0 No No This Windows Mobile comes with a JVM. MP3 Support: Yes No No A medium miss, since this is a relatively old device now 3GPP Support: Yes No No although it s still capable of at least H.263! Detected Browser: Opera Mobile 35166 Opera Mini 4 Not Available The change of browser from default really threw the other systems, but this isn t Opera Mini. Detected OS: Windows CE 5.1 Not Available Not Available Detected DP: Windows Mobile 5.0 Not Available Not Available Performance Mark: 5/5 0/5 0/5 A sidegrade to an old phone causes major ID problems.
Opera Windows Mobile Browser (WinMo x.x) Opera/9.80 (Windows Mobile; WCE; Opera Mobi/WMD-50369; U; ru) Presto/2.4.13 Version/10.00 An unidentifiable Windows Mobile Device running Opera Mobile: how will everyone cope? Detected As: Opera/9.80 (Windows Mobile; WCE; Opera Mobi/WMD-50369; U; ru) Presto/2.4.13 Version/10.00 (UserAgent) 276682458121546fd82d16c6d40f 3b66 Opera/9.80 (Windows Mobile; WCE; Opera Mobi/WMD-50369; U; ru) Presto/2.4.13 Version/10.00 Opera Mini 4 opera_mini_ver4_sub012168 Opera/9.80 (J2ME/MIDP; Opera Mini/4.0.12168/2444 Unknown Vendor Unknown Model Opera/9.80 (W Everyone failed on this, since there s no model to get. Opera Mini choice is a bit odd. Match here is to WML generic rather than generic_xhtml. Opera would probably have preferred XHTML. WURFL couldn t do anything with this. DeviceAtlas ran out of data at W. Screensize: 240x320 176x160 Not Available This is DR s generic fallback for Windows Mobile. It might well be wrong. MIDP version: No No No The true value of this is indeterminate. It s quite likely that this device has a JVM, but Windows Mobile is provided without one, and there s no WM version number to go on. MP3 Support: No No No Since WinMo supports Media player, DetectRight has slipped up here: however, it would also have checked accept strings to modify this value. 3GPP Support: No No No This is indeterminate, though again DR would have checked the accept strings for more clues. Detected Browser: Opera Mobile WMD-50369 Not Available Not Available Detected OS: Not Available Not Available Not Available WCE obviously stands for Windows CE, but DR missed this. Detected DP: Windows Mobile 0 DR gets WM right, but no version Performance Mark: 3/5 1/5 (got the browser manufacturer right) 0/5 Summary of detection performance (human generated)
HTC T7385 Touch Pro2 (WinMo 6.0) Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) PPC; 480x800; HTC_Touch_Pro2_T7385-Orange; OpVer 122.190.1.610 An advanced HTC device: should be straightforward! Detected As: HTC T7385 Touch Pro2 HTC Touch Diamond2 HTC Touch Diamond2 (T5353) WURFL fails on the ID. DeviceAtlas chooses a completely different HTC Touch, with drastic results. 510db7dda3e10a2e23fa3b772d958014 htc_touchdiamond2_ver1_subor ange Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) PPC; 480x800; HTC_Touch_Pro2_T7385-Orange; OpVer 122.190.1.610 Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) PPC; 480x800; HTC_Touch_Diamond2_T5353- Orange; OpVer 93.172.15.752 1694704 WURFL has fallen back to the MS Mobile Browser ID Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) PPC; 4 This is very odd. DeviceAtlas has stopped at the 4 in 480, and come up with a T5353 which it says has screensize 240 (not 480) x 800. It s wrong, but it s wrong interestingly! Screensize: 480x800 240x400 240x800 DeviceAtlas wrong, despite using the screen width in the UA as a guide. And WURFL is even more incorrect, again despite the UA giving the game away. MIDP version: No No No MP3 Support: Yes Yes Yes WURFL misses MP3 support thanks to conservative browser specs. (Mozilla 2.0 assumed). 3GPP Support: Yes Yes No It s odd that this device should support video but not MP3 in WURFL s output. DeviceAtlas is wrong here, and possibly for the Touch Diamond2 as well. Detected Browser: Mobile Internet Explorer 8.1 Microsoft Mobile Explorer Not Available WURFL flunks it here Detected OS: Windows CE 5.2 Windows Mobile OS 6.1 Windows 6.1 DA is mixing up OS and Developer Platform in its fields Detected DP: Windows Mobile 6.0 0 Developer Platform Performance Mark: 5/5 1/5 1/5 A mixed bag of fail, mostly.
HTC P4500 TyTN Touch Pro2 (WinMo 6.0) HTC_TyTN-Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; PPC) An oldie but a goodie: how will the systems react? Detected As: HTC P4500 TyTN HTC Hermes HTC TyTN II It s not a TyTN II. 56d4c31e011916aced79af8921 ef2971 HTC_TyTN-Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; PPC) htc_tytnii_ver1_subua 739515 HTC_TyTN-Mozilla/4.0 (compatible; MSIE 4.01; Windows CE; PPC; 240x320; HTC TyTN2) HTC_TyTN- What s odd about this is that DA detects this as the TyTN II based on the string TyTN. And then gets it wrong because of that assumption. Screensize: 240x320 240x320 240x320 TyTNs are similar enough for the data to be useful. MIDP version: 2.0 2.0 No Only DR knew about HTC s JVM. MP3 Support: Yes Yes Yes It can. 3GPP Support: Yes Yes No UAProfile says it can Detected Browser: Pocket Internet Explorer 4.1 Microsoft Mobile Explorer 6.12 Not Available What on earth is that browser version? Detected OS: Windows CE 3.0 Windows Mobile OS Windows Everybody gets this, only DR gets the version number though but gets it wrong Detected DP: Windows Mobile DR misses the Windows Mobile version here Performance Mark: 4.5/5 3/5 2/5 TyTN II isn t a great match, but it was OK
Samsung GT-i8000 (Omnia 2) (WinMo 6.1/6.5) SAMSUNG-GT-i8000Orange/NXXJA1 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) A nice Samsung device from Orange. How will the systems possibly cope? Detected As: Samsung GT-i8000 Generic XHTML Samsung GT-i8000 DeviceAtlas nails this one, as does DR. 8929d3990b97f33d7fc03450b4160bc d SAMSUNG-GT-i8000Orange/NXXJA1 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) generic_xhtml 1705173 Mozz SAMSUNG-GT-i Even though DeviceAtlas nailed this one, it s been a bit premature with the recognition, since this bit of UA string is shared with at least 7 other live Samsung models. There are more Uas from the i8000, so it worked out this time, but that was pretty much luck. Screensize: 480x800 128x92 480x800 Correct from DA. WURFL is still stuck in fallback land. MIDP version: 2.1 No 2.1 MP3 Support: Yes No Yes 3GPP Support: Yes No No It does actually support video, so DA has this wrong. Detected Browser: Mobile Internet Explorer 8.1 Not Available Not Available There is a lot of difference between MME and MobileIE! Detected OS: Windows CE 5.2 Not Available Windows Mobile 6.5 Professional Although DA is talking about a developer platform, not an OS as such. it got it right. Detected DP: Windows Mobile 6.0 Windows Mobile DR got this wrong. Performance Mark: 4/5 0/5 4/5 DA got the right result, but from a dubious heuristic.
RIM BlackBerry Devices BlackBerry user agents are not hugely challenging to a detection system. The challenge is in picking up alterations, and getting the OS right. You can generate more Windows Mobile detection tables like this at http://www.detectright.com/detector_inspector.html?request=blackberry
BlackBerry 9700 (RIM OS 5.0) BlackBerry9700/5.0.0.566 Profile/MIDP-2.1 Configuration/CLDC-1.1 VendorID/114 A great little juicy BlackBerry. Will the systems suck at it? Detected As: RIM BlackBerry 9700 RIM BlackBerry 9700 (Onyx) RIM BlackBerry 9700/Onyx Complete failure for WURFL. DA and DR nail it. fe8de02d1e25d706f4b75259c4cb 9fa6 blackberry9700_ver1 1669808 BlackBerry9700/5.0.0.566 Profile/MIDP-2.1 Configuration/CLDC-1.1 VendorID/114 BlackBerry9700/5.0.0. 207 Profile/MIDP-2.1 Configuration/CLDC- 1.1 VendorID/175 BlackBerry97 This seems presumptive from DA to use so few digits on the ID. This implies that they have no other models in the 97xx range, but it s unwise to assume there will never be. Screensize: 480x360 480x360 480x360 DA and DR are correct of course. MIDP version: 2.1 2.0 2.1 Even the User agent tells us this, but WURFL doesn t have a field for MIDP 2.1. MP3 Support: Yes Yes Yes Correct from DA and DR 3GPP Support: Yes Yes Yes Correct from DA and DR Detected Browser: RIM Browser 5.0 BlackBerry 4.0 Not Available DA doesn t contain browser information and WURFL missed the detection. Detected OS: RIM OS 5.0 RIM OS 4.0 RIM OS 5.0.0 DA and DR are on safe ground here. WURFL is a bit behind the times. Performance Mark: 5/5 4/5 5/5 DA Nails it until the 9777 comes along!
BlackBerry 8520 Curve (RIM OS 4.6) BlackBerry8520/4.6.1.314 Profile/MIDP-2.0 Configuration/CLDC-1.1 VendorID/118 Will this model serve the engines a curve-ball? Detected As: RIM BlackBerry 8520 RIM BlackBerry 8520 (Curve) RIM BlackBerry 8520/Curve WURFL s BlackBerry fallback works here for version 4, but it still doesn t get the model. eff177782e13eeb542ad3ce16d78 b008 blackberry8520_ver1 1563234 The record the data matches in the system. Generic IDs are not good. BlackBerry8520/4.6.1.314 Profile/MIDP-2.0 Configuration/CLDC-1.1 VendorID/118 BlackBerry8520/4.6.1.259 Profile/MIDP-2.0 Configuration/CLDC-1.1 VendorID/100 BlackBerry852 Not too bad from DA, but still reliant on BlackBerry not releasing any more 852x devices in the near future. Screensize: 320x240 320x240 320x240 DR and DA right: WURFL wrong. MIDP version: 2.0 2.0 2 DA and DR right. MP3 Support: Yes No Yes DA and DR right. 3GPP Support: Yes Yes Yes I think I m doing this wrong. Detected Browser: RIM Browser 4.6 BlackBerry 4.6.1 Not Available DR gets this right. Detected OS: RIM OS 4.6 RIM OS 4.6.1 RIM OS 4.6.1 DA and DR right: WURFL in the right ballpark. Performance Mark: 5/5 4/5 (MP3 wrong) 5/5 Summary of detection performance (human generated)
BlackBerry 8310 (RIM OS 4.2) BlackBerry8310/4.2.2 Profile/MIDP-2.0 Configuration/CLDC-1.1 VendorID/124 UP.Browser/5.0.3.3 An 8310, but not running BlackBerry Browser. Will anyone notice? Detected As: RIM BlackBerry 8310 RIM BlackBerry 8310 (Curve) RIM Blackberry 8310 DA and DR got this right c4542d3907190a6885343145d1d 3f4c5 blackberry8310_ver1_sub422 209247 BlackBerry8310/4.2.2 Profile/MIDP-2.0 Configuration/CLDC-1.1 VendorID/124 UP.Browser/5.0.3.3 BlackBerry8310/4.2.2 Profile/MIDP-2.0 Configuration/CLDC-1.1 VendorID/124 BlackBerry831 Again relying on an implied zero from DA. Screensize: 320x240 320x240 320x240 DA and DR correct MIDP version: 2.0 2.0 2 DR and DA correct. MP3 Support: Yes Yes Yes DR and DA correct. 3GPP Support: Yes Yes Yes DR and DA correct. Detected Browser: Openwave 5.0 BlackBerry 4.2.2 Not Available Only DetectRight noticed or cared that this isn t running a RIM browser. This might well have an implication for preferred markup. Detected OS: RIM OS 4.2 RIM OS 4.2.2 RIM OS 4.2.2 Everyone got this Performance Mark: 5/5 4/5 5/5
BlackBerry 9650 (RIM OS 5.0) BlackBerry9650/5.0.0.617 Profile/MIDP-2.1 Configuration/CLDC-1.1 VendorID/104 A regular BlackBerry. Detected As: RIM BlackBerry 9650 Device Match Failed RIM BlackBerry Tour 9630 DA missed this one because of its overly-optimistic digit matching. WURFL seems to have failed to fall back to BlackBerry 5.0. 2ab92e5913fe1dda6b201359f22d d415 BlackBerry9650/5.0.0.617 Profile/MIDP-2.1 Configuration/CLDC-1.1 VendorID/104 generic_xhtml 1657691 Mozz BlackBerry96 The overly ambitious digit matching does have some consequences later, since DA has detected an older device than it really is Screensize: 480x360 128x92 480x360 Though the screensize is right MIDP version: 2.1 No 2 The user agent tells you it s 2.1, so this is a pretty bad miss all round. It might be important to some app sellers. MP3 Support: No No Yes What is DR doing here? That s really strange. The device should support MP3. 3GPP Support: Yes No Yes DA and DR get this Detected Browser: RIM Browser 5.0 Not Available Not Available DR does it correctly. Detected OS: RIM OS 5.0 Not Available RIM OS 4.7.1 Only DR gets this right, since it s not ignoring evidence right in front of it. Performance Mark: 4/5 0/5 3/5
Apple iphone OS Devices ipod user agents were until recently a challenge to WURFL, but this problem has been fixed in the latest API iteration. The amount of apps available for iphone means that not all iphone accesses follow the Apple standard pattern. Other browsers that namecheckthe iphone would need to be checked for using deep user agent scanning. The detections here show that for standard iphone accesses, all of the engines perform adequately. You can generate more iphone OS detection tables like this at http://www.detectright.com/detector_inspector.html?request=apple
Apple iphone (iphone OS 3.1) Mozilla/5.0 (iphone; U; CPU iphone OS 3_1_3 like Mac OS X; de-de) AppleWebKit/528.18 (KHTML, like Gecko) Mobile/7E18 The iconic device of the millennium so far. Detected As: Apple iphone Apple iphone Apple iphone All correct 6eeb0707bcd564f39c91cc669df5 dd60 Mozilla/5.0 (iphone; U; CPU iphone OS 3_1_3 like Mac OS X; de-de) AppleWebKit/528.18 (KHTML, like Gecko) apple_iphone_ver1 205202 It s worrying that WURFL thinks this is a V1 Mozilla/5.0 (iphone; U; CPU like Mac OS X; en) AppleWebKit/420+ (KHTML, like Gecko) Version/3.0 Mobile/1A538a Safari/419.3 Mozilla/5.0 (iphone; U; C This is odd. WURFL has used a version 3.0 useragent, but the OS is down as 1.0. Equally, DeviceAtlas has the OS at 3.1.2, but the user agent used to detect it doesn t get as far as a version number: which means this is a guess. Screensize: 320x480 320x480 320x480 An iconic phone has an iconic screensize. With icons, of course. MIDP version: No No No Apple give Sun no break here. MP3 Support: Yes Yes Yes An entire economy depends on this. 3GPP Support: Yes Yes Yes and this Detected Browser: Apple WebKit 528.1 Safari Not Available Both DR and WURFL are correct: this is a reversal of the Android Webkit/Mobile safari incident. Detected OS: iphone OS 3.1 iphone OS 1.0 OSX 3.1.2 DA got this right, but only by coincidence. Performance Mark: 5/5 4/5 5/5 Everyone did OK
Apple ipod Touch (iphone OS 3.1) Mozilla/5.0 (ipod; U; CPU iphone OS 3_1_3 like Mac OS X; ja-jp) AppleWebKit/528.18 (KHTML, like Gecko) Mobile/7E18 (via babelfish.yahoo.com) Another iphone with a multilinguistic cheer this time, courtesy of Yahoo. Detected As: Apple ipod Touch Apple ipod Touch Apple ipod Touch Only the newest version of the WURFL API gets this right: previous versions would detect this as an iphone due to an inappropriate matching algorithm. 496b56a364fc314814b9f30de13a d3e9 Mozilla/5.0 (ipod; U; CPU iphone OS 3_1_3 like Mac OS X; ja-jp) AppleWebKit/528.18 (KHTML, like Gecko) Mobile/7E18 (via babelfish.yahoo.com) apple_ipod_touch_ver3_1_3 312415 WURFL makes this an iphone (see bug) Mozilla/5.0 (ipod; U; CPU iphone OS 3_1_3 like Mac OS X; xx-xx) AppleWebKit/528.18 (KHTML, like Gecko) Version/4.0 Mobile/7E16 Safari/528.16 Mozilla/5.0 (ipo If Apple ever release an ipoo then DA is in trouble. And I wouldn t put it past them, frankly. Screensize: 320x480 320x480 320x480 Correct MIDP version: No No No Correct MP3 Support: Yes Yes Yes Correct 3GPP Support: Yes Yes Yes Correct Detected Browser: Apple WebKit 528.1 Safari Not Available Detected OS: iphone OS 3.1 iphone OS 3.1.3 OSX 3.0 DA has an odd way of representing the iphone OS, and gets it wrong. Detected DP: iphone This is really OS, but DA s field structure doesn t differentiate between Apple OSX iphone style, and Apple OSX desktop style. Performance Mark: 5/5 5/5 4/5 Summary of detection performance (human generated)
Samsung Devices There are large numbers of varying Samsung devices: Android, Windows Mobile, proprietary, CDMA and GSM. The prolific nature of Samsung s release schedule means that devices are likely to be missing from device data engines Samsung s habit of prefixing devices and having marketing names means that representation of device names is often inconsistent within the standard databases. You can generate more Samsung detection tables like this at http://www.detectright.com/detector_inspector.html?request=samsung
Samsung GT-i7500 Galaxy (Android 1.5) Mozilla/5.0 (Linux; U; Android 1.5; ja-jp; Galaxy Build/CUPCAKE) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 Samsung s Galaxy phone has many UA variations: how is it handled? Detected As: Samsung GT-i7500 Samsung Galaxy Unknown Vendor Unknown Model 06d5da360bbd1d067ef2cd9e97ea 150b Mozilla/5.0 (Linux; U; Android 1.5; ja-jp; Galaxy Build/CUPCAKE) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 samsung_galaxy_ver1_es Mozilla/5.0 (Linux; U; Android 1.5; xx-xx; Galaxy Build/CUPCAKE) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 Mozilla/5.0 (Linux; U; Android 1.5; DR and WURFL get this DA gets to Android 1.5 and gives up when it encounters the j in ja-jp. Strange that it doesn t fall back to some kind of generic Android 1.5 dataset. Screensize: 320x480 320x240 Not Available DR gets this right. WURFL should get this right, having done the detection, but it still picking up Android SDK Screensizes MIDP version: No No No Who needs J2ME, mmh? MP3 Support: Yes Yes No DA wrong 3GPP Support: Yes Yes No DA wrong. Detected Browser: Mobile Safari 525.2 Android Webkit Not Available Detected OS: Android 1.5 Android 1.5 Not Available DA missed it Performance Mark: 5/5 5/5 0/5 WURFL up there, DA down there.
Samsung GT-B7610 (WinMo 6.1) Mozilla (Windows NT 5.1; U; en) SAMSUNG-GT-B7610 Opera/9.5 An easy useragent: or is it? Claiming to be Windows NT 5.1 and using Opera?. Detected As: Samsung GT-B7610 Opera 9 Unknown Vendor Unknown Model Only DR gets this right bf9cda47b60e2d15ec70a6a9 bb21dc5d Mozilla (Windows NT 5.1; U; en) SAMSUNG-GT-B7610 Opera/9.5 opera_9 Opera would be happy with WURFL s Opera/9 Mozilla ( WURFL missed most of the meat of the string, and DA gave up far too easily. Screensize: 480x800 800x600 Not Available DR gets this right. WURFL appears to think this is not a mobile device at all. Oops! This user might never see the premium mobile content. MIDP version: 2.1 No No Not in the UA, but present and correct. MP3 Support: Yes No No DR Correct 3GPP Support: Yes No No DR Correct Detected Browser: Opera 9.5 Opera 9 Not Available DR Correct Detected OS: Windows CE 5.2 Not Available Not Available DR Correct Detected DP: Windows Mobile 6.1 DR Correct Performance Mark: 5/5 1/5 0/5 WURFL gets two points for getting the browser right, but loses one for not detecting it as a mobile device.
Samsung GT-B5310 SAMSUNG-GT-B5310/B5310MMJC1 SHP/VPP/R5 Dolfin/1.5 Nextreaming SMM-MMS/1.2.0 profile/midp-2.1 configuration/cldc-1.1 A low-end Samsung device. If Dolfin is a browser then people had better start detecting it! Detected As: Samsung GT-B5310 Samsung GT-B5310 Samsung GT-B5310 WURFL back to generic 467d0738d72c31fd1bcfad0dc696e102 samsung_gt_b5310_ ver1 SAMSUNG-GT-B5310/B5310MMJC1 SHP/VPP/R5 Dolfin/1.5 Nextreaming SMM-MMS/1.2.0 profile/midp-2.1 configuration/cldc-1.1 SAMSUNG-GT- B5310/B5310ACIK1 SHP/VPP/R5 Dolfin/1.5 Nextreaming SMM- MMS/1.2.0 profile/midp-2.1 configuration/cldc -1.1 1817049 SAMSUNG-GT-B53 Screensize: 240x320 240x320 240x320 DA and DR are correct DA s presumptuous number matching makes it a hostage to fortune again, should there be a B5315. MIDP version: 2.1 2.0 No How did DA miss this? It s in the user agent, and we know they detected the correct device! WURFL again suffers from not having a MIDP2.1 field. MP3 Support: Yes Yes Yes DR and DA correct 3GPP Support: Yes No No DA and WURFL missed this one. Detected Browser: Not Available Not Available Not Available Unless the browser is Dolfin v 1.5, they all missed this. Detected OS: Not Available Not Available Not Available Proprietary OS. Performance Mark: 5/5 (no obvious browser) 4/5 (missed the video) 3/5 DeviceAtlas detects correctly, but lets itself down with two odd data results. WURFL misses video support.
Samsung SGH-D407 SEC-SGHD407 Old school Samsung Device. Should be easy, right? Detected As: Samsung SGH-D407 Samsung SGH-D407 Samsung SGH-D407 All OK f633c5d93dfd637473c1e3 2102c7835b sec_sghd407_ver1 209853 All OK SEC-SGHD407 SEC-SGHD407/1.0 TSS/2.5 SEC-SGHD407 All OK Screensize: 128x160 128x160 128x160 All OK MIDP version: 2.0 1.0 No DA and WURFL underestimate this dark horse of a phone! MP3 Support: Yes Yes No MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes Yes No Does this support 3GPP video (either H.263 or H.264)? Detected Browser: Not Available Not Available Not Available Actually someone should have detected TSS 2.5 here. Detected OS: Not Available Not Available Not Available Proprietary OS Performance Mark: 5/5 4/5 2/5 DA missed more data in this one
Sony Ericsson Devices Sony Ericsson devices do not tax the device identification algorithms much, since they have historically used tidy useragents with device information front-loaded: e.g. SonyEricssonZ550i/R6CA You can generate more Samsung detection tables like this at http://www.detectright.com/detector_inspector.html?request=sony%20ericsson
Sony Ericsson Z520i sonyericssonz520i/ucweb7.0.0.33/70/999 Another browser has got hold of this one. Will everyone detect it properly? Detected As: Sony Ericsson Z520i Device Match Failed Device Match Failed Nothing doing except for DR 8c1003a4a35deaa8ec40e6c7 d34e6b2e generic Interestingly, an additional test showed that Tera-WURFL detects this as a web browser, not a mobile device. sonyericssonz520i/ucweb7. 0.0.33/70/999 sonyericsson Well, at least DA knows it s from Sony Ericsson! Screensize: 128x160 90x40 Not Available DR Correct. MIDP version: 2.0 No No DR Correct. MP3 Support: Yes No No DR Correct. 3GPP Support: Yes No No DR Correct. Detected Browser: UCWEB 709 Not Available Not Available DR Correct. Detected OS: Not Available Not Available Not Available Proprietary OS Performance Mark: 5/5 0/5 0/5 Summary of detection performance (human generated)
Sony Ericsson K790i SonyEricssonK790i/R8BF Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0,sonyericssonk790i/UC Browser7.0.0.41/70/351 Another useragent hijacked by UCBrowser. Detected As: Sony Ericsson K790i SonyEricsson K790i Sony Ericsson K790i All systems got this right ab659d6ce70bb0b19d6adb77d1 4c7cf3 SonyEricssonK790i/R8BF Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0,sonyericssonk7 90i/UC Browser7.0.0.41/70/351 sonyericsson_k790i_ver1_subr1jc 204920 The record the data matches in the system. Generic IDs are not good. SonyEricssonK790i/R1JC Browser/NetFront/3.3 Profile/MIDP-2.0 Configuration/CLDC-1.1 SonyEricssonK790i This UA plays to DA s strengths. The WURFL UA picked out is slightly less of a good match because of the browser discrepency Screensize: 240x320 240x320 240x320 MIDP version: 2.0 2.0 2 All correct MP3 Support: Yes Yes Yes MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes Yes Yes Does this support 3GPP video (either H.263 or H.264)? Detected Browser: UNTRUSTED 1.0 Access Netfront 3.3 Not Available DR also failed to spot UCBrowser lurking on the end, but this shows its deductive reasoning picking UNTRUSTED/1.0 as the main brower (I suspect this access was made from a JVM then hijacked. Detected OS: Not Available Not Available Not Available Proprietary OS. More information from DR would be nice. Performance Mark: 4/5 3/5 3.5/5 (there s got to be a penalty there somewhere for not bothering with browsers!) Summary of detection performance (human generated)
Sony Ericsson W810i SonyEricssonW810i/R4EA Java/SEMC-Java/2.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0 A perfectly good SE phone but currently using Sony Ericsson s Java browser. Detected As: Sony Ericsson W810i SonyEricsson W810i Sony Ericsson W810i So far so good 75ac5b9d49cedb80dd18a8c97b3 6dc8b sonyericsson_w810i_subr4ea 205006 SonyEricssonW810i/R4EA Java/SEMC-Java/2.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0 SonyEricssonW810i/R4EA Browser/NetFront/3.3 Profile/MIDP-2.0 Configuration/CLDC-1.1 SonyEricssonW810i Once again WURFL has gone for a matching strategy which uses NetFront as the browser. That s not good. Screensize: 176x220 176x220 176x220 Screensize is very important. A failed detection can still get this right. MIDP version: 2.0 2.0 1 A bad miss for DA. MP3 Support: Yes Yes Yes MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes Yes Yes Does this support 3GPP video (either H.263 or H.264)? Detected Browser: SEMC Java 2.0 Access Netfront 3.3 Not Available This mistake by WURFL has possible implications for markup (DR would also interrogate the accept strings for help in a real-time situation). At least it gives browser information. Detected OS: Not Available Not Available Not Available Proprietary OS Detected DP: Not Available 0 JavaPlatform 6 DR should have this information, but doesn t. Performance Mark: 4/5 3/5 3/5 Summary of detection performance (human generated)
Sony Ericsson K800i SonyEricssonK800i/R1JE Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0;Opera/9.60 (J2ME/MIDP;Opera Mini/5.0.15650Mod.by.Handler/960; U;en-US)flameblur/2.2.0 A simple SE device hijacked by a modded Opera Mini. Could be trouble! Detected As: Sony Ericsson K800i SonyEricsson K800i Sony Ericsson K800i The classic SonyEricsson beginning to the string makes this an easy match 07e3ec46ec768502b1c03b27a78 3c491 SonyEricssonK800i/R1JE Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0;Opera/9.60 (J2ME/MIDP;Opera Mini/5.0.15650Mod.by.Handler/9 60; U;en-US)flameblur/2.2.0 sonyericsson_k800i_ver1_subr1jc 204922 SonyEricssonK800i/R1JC Browser/NetFront/3.3 Profile/MIDP-2.0 Configuration/CLDC-1.1 SonyEricssonK800i/ Again, the matching strategy for WURFL misses a crucial piece of the puzzle. Screensize: 240x320 240x320 240x320 Screensize is very important. A failed detection can still get this right. MIDP version: 2.0 2.0 2 OK MP3 Support: Yes Yes Yes MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes Yes Yes Does this support 3GPP video (either H.263 or H.264)? Detected Browser: Opera Mini 5.0 Access Netfront 3.3 Not Available Not knowing it s Opera Mini could have ramifications for content delivery. Detected OS: Not Available Not Available Not Available Proprietary OS Detected DP: Not Available JavaPlatform 7 DR should have this. Performance Mark: 4.5/5 (mark deducted for not having DP) 3/5 4.5/5 (DA sidestepped the difficult browser bit) Summary of detection performance (human generated)
Sony Ericsson X1i (WinMo 6.1) Mozilla (Windows NT 5.1; SonyEricssonX1i/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1; U) gzip(gfe) (via translate.google.com) Opera/9.50 A bit of a mess this one. First, the device ID is inside a Mozilla container: always bad for detection mechanisms. Secondly, we ve got a transcoder signature on the end! This is a pretty good phone, too. And notice it s claiming Windows NT 5.1 compatibility? Detected As: Sony Ericsson X1i Device Match Failed Unknown Vendor Unknown Model 73d3367eec6fc04f37028f716705c 987 Mozilla (Windows NT 5.1; SonyEricssonX1i/1.0 Profile/MIDP- 2.0 Configuration/CLDC-1.1; U) gzip(gfe) (via translate.google.com) Opera/9.50 generic_xhtml Only DR gets this right. Mozz Mozilla ( DA s tree strategy has an epic fail before getting to the telltale data, and WURFL falls straight through. Screensize: 480x800 128x92 Not Available DR Correct MIDP version: 2.0 No No DR Correct. MP3 Support: Yes No No DR Correct 3GPP Support: Yes No No DR Correct (no, seriously) Detected Browser: Opera 9.5 Not Available Not Available Because DR does deep substring analysis, it doesn t matter where Opera appears in the UA. Detected OS: Windows CE 19915.1 Not Available Not Available DR got this from its contains data: lists of components shipped with devices which are merged or swapped out in real-time with other detected components. Detected DP: Windows Mobile 6.1 0 Developer Platform Performance Mark: 5/5 0/5 0/5 Summary of detection performance (human generated)
Sony Ericsson K550i (Symbian) FIREFOX_SPEEDY(compatible; MSIE 6.0; Symbian OS; SonyEricsson K550i/R6B/5.0737.3.0.1; 9399) Opera 9.85 [en] >> You see some strange useragents on the Internet, and this one is pretty odd. Detected As: Sony Ericsson K550i SonyEricsson P800 Unknown Vendor Unknown Model 751d63f90936ff7cf3b5a1758 362f21b FIREFOX_SPEEDY(compatible; MSIE 6.0; Symbian OS; SonyEricsson K550i/R6B/5.0737.3.0.1; 9399) Opera 9.85 [en] >> opera_sonyericsson_p800_ver 1_sub630 Mozilla/4.0 (compatible; MSIE 5.0; Symbian OS; SonyEricsson P800; 305) Opera 6.30 [en] F DR gets this right. Interestingly, Tera- WURFL misses this entirely, making it a generic xhtml. The record the data matches in the system. Generic IDs are not good. F?? F??? At least WURFL s matching algorithm had a try!! Interesting that DA doesn t even have the string Firefox in it Screensize: 176x220 208x320 Not Available Too big by WURFL. MIDP version: 2.0 2.0 No Somehow I don t think this UA owner is a huge app buyer, but you never know MP3 Support: Yes Yes No Yep 3GPP Support: No Yes No Not a video phone. Detected Browser: Opera 9.8 Opera 6.30 Not Available Right browser, wrong version. Which is still pretty good. Detected OS: Symbian OS Symbian OS Not Available Detected OK by DR and WURFL. Performance Mark: 5/5 2.5/5 (device epic fail, marks for browser and DP) 0/5 Summary of detection performance (human generated)
Vodafone devices Vodafone has its own useragent structure it likes to follow, but it s also inconsistent across time, even within useragents for the same model. Vodafone s structuring of the useragent string makes it very difficult for substring matches based on the beginning of the useragent, and also makes least distance detection problematic. You can generate more Samsung detection tables like this at http://www.detectright.com/detector_inspector.html?request=vodafone
Samsung GT-B2100 (Vodafone) Vodafone/1.0/SAMSUNG-B2100V/BUIA2/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 UP.Browser/6.2.3.3.c.1.101 (GUI) MMP/2.0 Vodafone do a lot of mucking about with their UAs. Detected As: Samsung GT-B2100 Samsung B2100 Samsung B2100 The official name of this device is GT- B2100. Device catalogues often have some devices with the prefix and some without, which is messy and appears unprofessional in a pick list. 10fa685ebfee1d26c9d58c0cba09 8068 Vodafone/1.0/SAMSUNG- B2100V/BUIA2/1.0 Profile/MIDP- 2.0 Configuration/CLDC-1.1 UP.Browser/6.2.3.3.c.1.101 (GUI) MMP/2.0 samsung_b2100v_ver1 1348501 Pretty exact match here. Vodafone/1.0/SAMSUNG- B2100V/BUIA2/1.0 Profile/MIDP- 2.0 Configuration/CLDC-1.1 UP.Browser/6.2.3.3.c.1.101 (GUI) MMP/2.0 Vodafone/1.0/SAMSUNG-B DeviceAtlas is lucky here that its own internal problem space this time matches the actual problem space: there aren t any more Bxxxx models from Samsung sharing this UA structure. WURFL seems to have seen this UA before. Screensize: 128x160 128x160 128x160 Screensize is very important. A failed detection can still get this right. MIDP version: 2.0 No No Bad MIDP miss there. MP3 Support: Yes Yes Yes MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes No Yes WURFL has this wrong. Detected Browser: Openwave 6.2 Openwave Mobile Browser 6.2 Not Available All present and correct. It s understandable why DA leaves browser out of it: their matching algorithm would make the results a complete mess. Detected OS: Not Available Not Available Not Available Proprietary OS Performance Mark: 5/5 4/5 5/5 (but no browser info) Judging by the browser age, this UA has been around a while.
Vodafone V1240 VodafoneV1240/B037 Browser/NetFront/3.5 MMS/Obigo-MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/2.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Player/QTV-Player/5.3 There s a lot of information here in this UA how will the engines cope? Detected As: Vodafone v1240 Huawei V1240 Unknown Vendor Unknown Model 51704e9c58cf13d4cdef98a4b618 12ca VodafoneV1240/B037 Browser/NetFront/3.5 MMS/Obigo- MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/2.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Player/QTV- Player/5.3 huawei_v1240_ver1 VodafoneV1240/B037 Browser/NetFront/3.5 MMS/Obigo- MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/2.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Player/QTV- Player/5.3 Vodafone Screensize: 240x320 240x320 Not Available DA fails. MIDP version: 2.0 2.0 No DA fails. MP3 Support: Yes Yes No DA fails. There s an interesting point here, which is that this device is both of these things. DetectRight has a big list of AKAs, which also has WURFL s main name in it for this device. We called it a Vodafone V1240 because Vodafone sell it as a Vodafone branded device, and it s only made by Huawei: though they also sell it under their own name with a different UA. Not much luck with DA, but WURFL has something. 3GPP Support: No No No This must be wrong. This device has a QTV player in it, which ought to handle video. I think all the systems are wrong. Detected Browser: NetFront 3.5 Access Netfront 3.5 Not Available DR and WURFL correct Detected OS: Windows CE 5.1 Windows Mobile OS 5 Not Available Detected DP: Windows Mobile 5.0 0 Developer Platform Performance Mark: 4/5 4/5 0/5 Summary of detection performance (human generated)
LG HB620T Vodafone/1.0/LG-HB620T/V10a Browser/Obigo-Q05A/3.12 MMS/LG-MMS-V1.0/1.2 Java/ASVM/1.1 Profile/MIDP-2.1 Configuration/CLDC-1.1 An LG model lurking in here. Will it get spotted? Detected As: LG HB620T LG KU990 LG HB620T WURFL guesses wrong based on very similar useragents: similar enough to confuse the LD algorithm and make it ditch the model name. 2f893042193829773e9fdf5d7e50 bfc1 Vodafone/1.0/LG-HB620T/V10a Browser/Obigo-Q05A/3.12 MMS/LG-MMS-V1.0/1.2 Java/ASVM/1.1 Profile/MIDP-2.1 Configuration/CLDC-1.1 lg_ku990_ver1_subv10c 1383494 Vodafone/1.0/LG-KU990/V10c Browser/Obigo-Q05A/3.6 MMS/LG-MMS-V1.0/1.2 Java/ASVM/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Vodafone/1.0/LG-H DA s strategy worked this time to be fair, if other models came along they knew about, the tree would split properly at another point. But it depends on DA knowing about it. WURFL s distance match has been overwhelmed by the similarities between the useragents, but missed the vital model information. Screensize: 320x240 240x400 320x240 WURFL gets this wrong, since it thinks this is another device. MIDP version: 2.1 2.0 2.1 I m sure WURFL would put 2.1 if it had a 2.1 field. MP3 Support: Yes Yes Yes OK 3GPP Support: Yes Yes No No video support of any kind, DA? Isn t that a bit mean? Detected Browser: Obigo Q05A Teleca-Obigo 5.0 Not Available Both good choices. Performance Mark: 5/5 2/5 (Model is wrong, markup would probably have worked) 4/5 DR and DA get this one.
Vodafone 725 Vodafone/1.0/0Vodafone725/B114 Browser/Obigo-Browser/Q05A MMS/Obigo-MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Player/QTV-Player/5.3 UNTRUSTED/1.0 Vodafone do a lot of mucking about with their UAs. Detected As: Vodafone 725 Huawei Vodafone 725 Huawei V720 WURFL s result shows how difficult it is to stay consistent with model names when you ve got multiple parties and/or marketing names involved. DA got it actually badly wrong. bd8b19b04fab81be89af6cd229081a9d huawei_vf725_ver1_subvodaf one10 Vodafone/1.0/0Vodafone725/B114 Browser/Obigo-Browser/Q05A MMS/Obigo- MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Player/QTV-Player/5.3 UNTRUSTED/1.0 Vodafone/1.0/0Vodafone725/B1 12 Browser/Obigo- Browser/Q05A MMS/Obigo- MMS/Q05A SyncML/HW- SyncML/1.0 Java/HWJa/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 Player/QTV-Player/5.3 649895 Vodafone/1.0/0Vodafone72 Screensize: 240x320 240x320 176x220 DA wrong. MIDP version: 2.0 2.0 2 Everyone got this Device Atlas s premature user agent matching bites it in the rear end here, since the phone it s detected is substantially different to the one we re seeing. MP3 Support: Yes Yes Yes MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes Yes No Misdetection costs opportunities Detected Browser: Obigo Q05A Teleca-Obigo Q05A Not Available This one raised my Obigo brow Detected OS: Not Available Not Available BREW 3.1 Java and BREW in the same device? BREW is more of a developer platform than an OS (it s normally coupled with REX OS). Performance Mark: 5/5 5/5 2/5 DA shows its limitations here
Vodafone 810 Vodafone-810/1.0 Browser/Obigo-Browser/Q05A MMS/Obigo-MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/1.0 Profile/MIDP-2.0 Configuration/CLDC-1.1 UNTRUSTED/1.0 A newer Vodafone, but medium spec. How will the chaps do? Detected As: Vodafone 810 Device Match Failed Unknown Vendor Unknown Model 403ec772ae9b21d896dad36e14318953 generic DR Correct Vodafone-810/1.0 Browser/Obigo-Browser/Q05A MMS/Obigo-MMS/Q05A SyncML/HW-SyncML/1.0 Java/HWJa/1.0 Profile/MIDP-2.0 Configuration/CLDC- 1.1 UNTRUSTED/1.0 Vodafone DR Correct Screensize: 240x320 90x40 Not Available DR Correct MIDP version: 2.0 2.0 No DR Correct MP3 Support: Yes No No DR Correct 3GPP Support: Yes No No DR Correct Detected Browser: Obigo Q05A Not Available Not Available DR Correct Detected OS: Not Available Not Available Not Available DR Correct. We could probably do better with the OS though. Performance mark: 5/5 0/5 0/5 DR Correct
Verizon Phones Verizon are fond of Mozilla formatted useragents Verizon s Android naming with HTC-sourced devices means that names can get pretty complicated. You can generate more Verizon detection tables like this at http://www.detectright.com/detector_inspector.html?request=verizon
Verizon ADR6200 (Eris/Desire) Mozilla/5.0 (Linux; U; Android 1.5; es-us; ADR6200 Build/CUPCAKE) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 A lovely phone, but will it be underappreciated? Detected As: Verizon ADR6200 HTC Desire ADR6200 HTC Desire ADR6200 Right model number but misses the Verizon bit. DR has the other names as AKAs. 13fbff9072cb4257e43a49e2cb14 8fd5 Mozilla/5.0 (Linux; U; Android 1.5; es-us; ADR6200 Build/CUPCAKE) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 htc_eris_ver1_subadr6200 1842680 Mozilla/5.0 (Linux; U; Android 1.5; en-us; ADR6200 Build/CUPCAKE) AppleWebKit/528.5+ (KHTML, like Gecko) Version/3.1.2 Mobile Safari/525.20.1 Mozilla/5.0 (Linux; U; Android 1.5; es-us; A DA manages a correct analysis, and WURFL has seen this UA before. But Screensize: 320x480 320x240 320x240 What happened here? MIDP version: No 2.0 No We don t think Verizon wanted Java on this MP3 Support: Yes Yes Yes OK 3GPP Support: Yes Yes No DA leaves this Android user undersized and understimulated.. Detected Browser: Mobile Safari 525.2 Android Webkit Not Available Detected OS: Android 1.5 Android 1.5 Android Correct (amundo). Performance Mark: 5/5 3/5 2/5 Screensize and Java are the problem here for WURFL. DA compounds the problem with video support missing.
Verizon ADR6300 (Incredible!) Mozilla/5.0 (Linux; U; Android 2.1-update1; en-us; ADR6300 Build/ERE27) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 An impressive name, but will the detection live up to the name? Detected As: HTC Droid Incredible HTC ADR6300 HTC 6277 A mixed bag, but WURFL and DR are on the same page. What on earth happened to DA? 785c67d10b327973d68062716b3 74377 Mozilla/5.0 (Linux; U; Android 2.1- update1; en-us; ADR6300 Build/ERE27) AppleWebKit/530.17 (KHTML, like Gecko) Version/4.0 Mobile Safari/530.17 htc_incredible_adr6300_ver1 1689185 Mozilla/5.0 (Linux; U; Android 2.1- update1; en-us; ADR6300 Build/ERE27) AppleWebKit/525.10 (KHTML, like Gecko) Version/3.0.4 Mobile Safari/523.12.2 Mozilla/5.0 (Linux; U; Android 2.1- update1; en-us; A How did DA get to HTC 6277 from that string? Screensize: 480x800 480x800 320x480 DA underpowers the device. Let s hope the owner has good eyesight. MIDP version: No 2.0 No This is a Verizon machine, so MIDP is off the menu. MP3 Support: Yes Yes Yes MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes Yes No Does this support 3GPP video (either H.263 or H.264)? Detected Browser: Mobile Safari 530.1 Android Webkit Not Available The browser detected from this UA. It should be obvious if this is correct. Detected OS: Android 2.1 Android 2.1 Android DMSS7600/Cupcake DA wrong, since the model is wrong. Performance Mark: 5/5 4/5 1/5 Summary of detection performance (human generated)
UT Starcom XV6800 (WinMo 6.0) 320x320; VZW; UTStar-XV6800; Window Mobile 6.0 Professional; (compatible; MSIE 6.0; Windows CE; IEMobile 6.12) 320x320; VZW; UTStar-XV6800; Window Mobile 6.0 Professional; A weird but real useragent, lying about its screensize! Detected As: UTStarcom XV6800 Generic Windows Mobile Unknown Vendor Unknown Model b51d5b2a16cd7411b345195 8a3097dcb 320x320; VZW; UTStar- XV6800; Window Mobile 6.0 Professional; (compatible; MSIE 6.0; Windows CE; IEMobile 6.12) 320x320; VZW; UTStar-XV6800; Window Mobile 6.0 Professional; ms_mobile_browser_ver1 Mozilla/2.0 (compatible; MSIE XXXXXX; Windows CE; Smartphone; XXXXXXX) Screensize: 240x320 176x220 Not Available DR Correct Only DR gets this one. WURFL matches this to a very generic Windows Smartphone. A smarter generic these days would be 240x320, I think. MIDP version: 2.0 2.0 No It s odd that WURFL s generic Windows Mobile has MIDP, since WinMo doesn t automatically come with a JVM. Seems like J2ME got in here since BREW doesn t play well with WinMob. MP3 Support: Yes No No DR Correct 3GPP Support: Yes No No DR Correct Detected Browser: Mobile Internet Explorer 6.12 Microsoft Mobile Explorer Not Available DR Correct. MME is a very different beast from IEMobile. Detected OS: Windows CE 5.2 Windows Mobile OS Not Available DR gets the version right. Detected DP: Windows Mobile 6.0 Developer Platform Performance Mark: 5/5 1/5 0/5 Summary of detection performance (human generated)
Sprint Phones Sprint is a telcowho often rebrands devices based on a common hardware platform: especially with HTC devices. Sprint has its own java extensions library for J2ME Sprint is connected to Bell Mobile in Canada You can generate more Sprint detection tables like this at http://www.detectright.com/detector_inspector.html?request=sprint
Sprint Touch Pro 2 (T7380) Sprint/1.0/HTC_Touch_Pro2_T7380 Opera/9.70 (Windows NT 5.1; U; en) Will this great phone get a poor reception? Detected As: Sprint Touch Pro 2 opera 9 HTC PPC6900 WURFL thinks this is a normal Opera device. DA s match is eccentric. a778468ad81b6a76c35fcf5272 ad2492 Sprint/1.0/HTC_Touch_Pro2_T7 380 Opera/9.70 (Windows NT 5.1; U; en) opera_9 1113269 Opera/9 Spr DA s match is explained: it used only three letters. So this device would also be detected for any other device with a Sprint/1.0 useragent. Or indeed the poor Spring 5500 which doesn t have a prayer. Screensize: 480x800 800x600 320x240 Proper detection leads to proper screensizes. MIDP version: 2.0 No 2 Opera web doesn t have J2ME of course. Internally consistent, but wrong in context. MP3 Support: Yes No Yes Some accidental similarities between HTC models means DA is accurate here. 3GPP Support: Yes No No Only DR is right. Detected Browser: Opera 9.7 Not Available Not Available It s strange that WURFL doesn t have an entry here, considering that it s a browser entry. Detected OS: Windows CE 5.2 Not Available Windows Detected DP: Windows Mobile 6.1 0 Performance Mark: 5/5 1/5 1/5 (gets MP3 support right) DA not keeping up with the global expanding world of useragents.
Sprint PPC6900 Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 6.12) Sprint:PPC6900 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) Sprint MP6900SP An extremely odd useragent containing two model names, almost like two systems fighting it out. Remember this is the kind of stuff detection systems in the wild have to deal with. Detected As: Sprint PPC6900 Generic Windows Mobile HTC PPC6800 WURFL mishits, and DA chooses a previous model, this time with less ill effects. df898ef3bd542afde06f1f34cca21 5fa Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 6.12) Sprint:PPC6900 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) Sprint MP6900SP ms_mobile_browser_ver1 312475 Mozilla/2.0 (compatible; MSIE XXXXXX; Windows CE; Smartphone; XXXXXXX) Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 6.12) Sp Screensize: 240x320 176x220 240x320 Screensize is very important. A failed detection can still get this right. MIDP version: 2.0 No 2 WURFL misses out MP3 Support: Yes No Yes 3GPP Support: Yes No No DA wrong here, thanks to the device chosen being older than the device seen. Detected Browser: Mobile Internet Explorer 6.1 Microsoft Mobile Explorer Not Available We could have chosen either browser version here. Detected OS: Windows CE 5.2 Windows Mobile OS Windows 6.0 Mostly right. Detected DP: Windows Mobile 6.0 Performance Mark: 5/5 1/5 3/5 Summary of detection performance (human generated)
Sprint PPC6800 Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) Sprint:PPC6800,gzip(gfe) (via translate.google.com) An extremely odd useragent, like two systems fighting it out. Detected As: Sprint PPC6800 Generic Windows Mobile HTC PPC6800 DA and DR get this right 6cbc7345e7d0a94c9e69dd76627 764e9 Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) Sprint:PPC6800,gzip(gfe) (via translate.google.com) ms_mobile_browser_ver1 312475 Mozilla/2.0 (compatible; MSIE XXXXXX; Windows CE; Smartphone; XXXXXXX) Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) Sprint:P Usual comments apply to DA s detection, but DR knows of no other user agents with this prefix either. Screensize: 240x320 176x220 240x320 DA and DR have the right device. MIDP version: 2.0 No 2 MP3 Support: Yes No Yes 3GPP Support: Yes No Yes Detected Browser: Mobile Internet Explorer 7.11 Microsoft Mobile Explorer Not Available Generic browser reference from WURFL Detected OS: Windows CE 5.2 Windows Mobile OS Windows Generic OS reference from WURFL Detected DP: Windows Mobile 6.0 DR really gets it. Performance Mark: 5/5 1/5 5/5 DA and DR are on the same page.
Sprint PPC6850SP Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) Cricket PPC6850SP Will this be detected, or isn t it cricket? Detected As: Sprint PPC6850 Generic Windows Mobile Unknown Vendor Unknown Model 90383477e7faf4af6f29d11ca204 334d ms_mobile_browser_ver1 DR sees this, WURFL goes generic. DA gives up. Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) Cricket PPC6850SP Mozilla/2.0 (compatible; MSIE XXXXXX; Windows CE; Smartphone; XXXXXXX) Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) DA tried hard this time, but gave up at Cricket. Screensize: 480x640 176x220 Not Available MIDP version: 2.0 No No Generic detections just aren t cricket. MP3 Support: Yes No No You know Yes would be a safer assumption these days. 3GPP Support: Yes No No Detected Browser: Mobile Internet Explorer 8.1 Microsoft Mobile Explorer Not Available DR bowled a Googly here Detected OS: Windows CE 5.2 Windows Mobile OS Not Available Useful information from DR Detected DP: Windows Mobile 6.0 Performance Mark: 5/5 1/5 0/5 Other device detection systems get stumped.
Sprint MP6950SP/BM Mozilla (Windows Mobile; HTC-MP6950 BM; Opera Mobi/35166; U; en) Presto/2.2.1 Opera/9.7 Sideloaded Opera Mobile causes problems on this smartest of smartphones. Detected As: Sprint MP6950SP Device Match Failed Unknown Vendor Unknown Model 86a8c492c02388a48d0f27db 73ea6f79 Mozilla (Windows Mobile; HTC-MP6950 BM; Opera Mobi/35166; U; en) Presto/2.2.1 Opera/9.7 generic_xhtml Screensize: 480x640 128x92 Not Available MIDP version: 2.0 No No MP3 Support: Yes No No The sidegrade to Opera Mobile really threw off the detection. This illustrates the difficulty static device data based on UA catchup has in dealing with the behaviour of real users. Mozz Mozilla ( A lot of things can happen to a device after it s shipped, and many of these show up in the user agent. Deep analysis of the UA string is pretty much the only way to go if you want to catch things first time with any specificity. 3GPP Support: Yes No No Detected Browser: Opera Mobile 35166 Not Available Not Available We re in Opera Mobile territory. Detected OS: Windows CE 5.2 Not Available Not Available The OS detected from this UA. It's crucial to get this right. Detected DP: Windows Mobile 6.1 0 Developer Platform Performance Mark: 5/5 0/5 0/5
O2 Devices O2 also has many rebranded devices based on existing hardware platforms, initially rebranding Maxon devicies and progressing to HTC. The XDA model brand is unique to this telco. You can generate more Samsung detection tables like this at http://www.detectright.com/detector_inspector.html?request=o2
O2 XDA Atom Xda Atom Profile/MIDP-2.0 Configuration/CLDC-1.1 Will detection performance be Atomic? Or sub-atomic? Detected As: O2 XDA Atom Device Match Failed O2 XDA Orbit DA has the wrong XDA. fbf411092bac1265dacb4f4 a8171ed7d Xda Atom Profile/MIDP- 2.0 Configuration/CLDC- 1.1 generic 985075 Xda DA will match all O2 XDA devices to the Orbit. Screensize: 240x320 90x40 240x320 Standard Windows QVGA MIDP version: 2.0 No No MP3 Support: Yes No Yes 3GPP Support: Yes No No Detected Browser: Pocket Internet Explorer 4 Not Available Not Available An integral default part of Windows Mobile 5.0 Detected OS: Windows CE 5.1 Not Available Not Available Detected DP: Windows Mobile 5.0 0 Performance Mark: 5/5 0/5 2/5 WURFL misses out, and DA is only right when it is by virtue of product family.
O2 XDA Flint Xda_Flint/640x480 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) Is this detection hard as flint? Or easy as pie? Detected As: O2 XDA Flint O2 Xda Flint Unknown Vendor Unknown Model 274bece21e3fccc5049d9127f6 c673e9 Xda_Flint/640x480 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) htc_xda_flint Xda_Flint/670x480 (compatible; MSIE 6.0; Windows CE; IEMobile 7.11) Xda_ WURFL gets this right, as does DR. DA gets to the underscore and gives up. I think DA was probably able to work out that this was a mobile device, but that s probably about it. Screensize: 640x480 240x320 Not Available WURFL messed up the screensize here. And isn t that a weird screensize I the user agent? MIDP version: 2.0 2.0 No Everyone except DA on a MIDP 2.0 binge. MP3 Support: Yes Yes No 3GPP Support: Yes Yes No Detected Browser: Mobile Internet Explorer 7.1 Microsoft Mobile Explorer 6.12 Not Available I m not sure about WURFLs naming of Microsoft Browsers. They re not all MME, some of them are Pocket Internet Explorer, and some IE Mobile. They have different characteristics. Also, WURFL s version number is wrong. Detected OS: Windows CE 5.2 Windows Mobile OS 5 Not Available WURFL must be falling back to get this Detected DP: Windows Mobile 6.0 0 Performance Mark: 5/5 3.5/5 0/5 Another miss for DA. XDA is not its strongest point.
O2 XDA Exec Opera/9.5 (Microsoft Windows; PPC; Xda Exec; 640x480; Opera Mobile/405; U; en) Which executive functions will get it right? Detected As: O2 XDA Exec HTC P3702 Unknown Vendor Unknown Model 488233360a0a803febc52dde72278 3b0 htc_diamond_ver1_3702victor_sub en This is an odd WURFL result. I ll explain why further down. Opera/9.5 (Microsoft Windows; PPC; Xda Exec; 640x480; Opera Mobile/405; U; en) Opera/9.5 (Microsoft Windows; PPC; Opera Mobi/1938; U; en) Opera/9.5 (Microsoft Windows; PPC; DA gave up just before the payload. With WURFL, this is an example of something it does quite regularly: overinterpret a user agent. In this case, that user agent cannot be specifically tied to the HTC P3702: it s generic. OK, a P3702 was almost certainly seen with this user agent (and they know it was a P3702 because of the UAProfile URL), but that doesn t mean you can make this assumption. Screensize: 640x480 480x640 Not Available Presume DR is right since it got the right device. MIDP version: 2.0 2.0 No MP3 Support: Yes Yes No WURFL s choice isn t too bad in this case, since they re both high powered HTC-sourced devices. 3GPP Support: Yes Yes No Detected Browser: Opera Mobile Opera Mini 7.11 Not Available Opera Mobile is not the same as Opera Mini. Opera Mobile is a client-based browser on highend phones, Opera Mini is a proxy-based browser for most MIDP2 phones (and now other smartphones). Detected OS: Windows CE 14331.0 Windows Mobile OS 6.1 Windows XDA Exec was the first XDA to run Windows 6.0. DR s CE seems to be the build version though. Detected DP: Windows Mobile 6.0 0 Developer Platform Performance Mark: 5/5 2/5 1/5 (gets Windows CE right) Summary of detection performance (human generated)
O2 XDA II Mini Mozilla/4.0 (compatible; MSIE 4.01; Windows CE; O2 XDA II mini; 240x320) Is small beautiful for these detection engines? Or will this poor device be overlooked? Detected As: O2 XDA II Mini Generic Windows Mobile Unknown Vendor Unknown Model 73faffdc60d9ab62aa4023098 de4763d Mozilla/4.0 (compatible; MSIE 4.01; Windows CE; O2 XDA II mini; 240x320) ms_mobile_browser_ver1 Mozilla/2.0 (compatible; MSIE XXXXXX; Windows CE; Smartphone; XXXXXXX) Mozilla/4.0 (compatible; MSIE 4.01; Windows CE; O2 X DR gets this, WURFL has a stab but chooses the wrong branch of the WinMob tree (SmartPhone branch as opposed to PPC branch). DA s strategy is puzzling again here. Screensize: 240x320 176x220 Not Available DR correct. The resolution is also in the user agent. MIDP version: 2.0 No No DR correct MP3 Support: Yes No No DR correct 3GPP Support: Yes No No DR correct Detected Browser: Pocket Internet Explorer 4 Microsoft Mobile Explorer Not Available I prefer the PIE designation Detected OS: Windows CE Windows Mobile OS Not Available Oddly DA had this information (it knows this is Windows CE) but chose not to use it. Detected DP: Windows Mobile 2003 SE Performance Mark: 5/5 1/5 0/5
KDDI/AU Devices KDDI/AU devices feature codes in their user agents, which have to be looked up online. Unlike many other useragents, it s not immediately obvious which device is which from the code. Since devices in these ranges are co-produced between an OEM and KDDI, they end up with many different name variants: for instance, Toshiba W64T, KDDI W64T, KDDI-Toshiba W64T. DR contains all of these variants to help searching. These devices don t support MP3, although most now can support 3GP2. This appears to be a policy decision to lock users into purchasing the SMAF format. QCELP is the main free audio format supported. You can generate more KDDI/AU detection tables like this at http://www.detectright.com/detector_inspector.html?request=kddi/au
Toshiba/KDDI/AU PLY Mozilla/5.0 Opera/9.5 (KDDI-TS3Q; BREW; Opera Mobi; U; ja) Presto/2.2.1 A powerful Toshiba phone, with an interesting story, told in layers. Detected As: Toshiba PLY Device Match Failed Unknown Vendor Unknown Model 8ad88d120c9e980aebfacd362 0bc3a46 Mozilla/5.0 Opera/9.5 (KDDI- TS3Q; BREW; Opera Mobi; U; ja) Presto/2.2.1 opera Complete detection failure for DA and WURFL, despite containing this device in their data. This illustrates the importance of detection in matching data. Opera Mozilla/5.0 The combination of the Mozilla and the Opera strings in succession appears to have done the damage here. Screensize: 480x854 800x600 Not Available Strangely, treating this as a web browser probably isn t too bad, but it mucks up your mobile analytics. MIDP version: No No No Device has no need of J2ME, for it has BREW MP3 Support: No No No Sound has to be embedded in SMAF. This is more a policy than a technical issue I guess. 3GPP Support: Yes No No Well, 3GP2 support, technically. Detected Browser: Opera Mobile Not Available Not Available Shouldn t this say Opera for WURFL? Detected OS: Not Available Not Available Not Available DR really needs to get under the skin of these AU phones a little more. Detected DP: BREW Performance Mark: 5/5 1.5/5 (would target web rather than mobile site) 0/5 Both DA and WURFL thrown off by the irregular string.
Kyocera/KDDI W65K Mozilla/4.0 (compatible; MSIE 6.0; KDDI-KC3I) Opera 8.60 [en] A Kyocerian Opera awaits. What performance can we expect? Detected As: Kyocera W65K kddi W65K Unknown Vendor Unknown Model 30eb26983e51de9e85585066 99c1f7ea Mozilla/4.0 (compatible; MSIE 6.0; KDDI-KC3I) Opera 8.60 [en] kddi_kc3i_ver1_submozilla Mozilla/4.0 (compatible; MSIE 6.0; KDDI-KC3I) Opera 8.60 [en] Mozilla/4.0 (compatible; MSIE 6.0; DR also has the KDDI aka attached to this model. DA thought this was a web browser, I think. Screensize: 240x400 240x400 Not Available Two good, one bad/ugly. MIDP version: No No No EZAppl/BREW rather than J2ME for this CDMA phone. MP3 Support: No No No MP3 supported? Mistakes here could mean customers missed. 3GPP Support: Yes No No Technically supports 3GP2 files, but supports the H263 codec within that. Detected Browser: Opera 8.6 Openwave Mobile Browser 6.2 Not Available This is a strange result from WURFL considering the string it chose. Detected OS: Not Available Not Available Not Available Ah, mysterious operating systems Performance Mark: 5/5 3.5/5 0/5 WURFL misses out on browser and video. DA just misses.
Nokia Devices Historically Nokia Devices have been fairly easily detected, except for short periods when the browser department decides that device ID isn t necessary (see: Series 60 3.0, a policy decision now dismissed as a bug). The move to Mozilla has caused more trouble for detection engines, which rely on the ubiquity of Nokia devices to get enough user agents to keep up to date. You can generate more Nokia detection tables like this at http://www.detectright.com/detector_inspector.html?request=nokia
Nokia E72 (Series 60 3.2) NokiaE72/021.024 (SymbianOS/9.3; U; Series60/3.2 Mozilla/5.0; Profile/MIDP-2.1 Configuration/CLDC-1.1) AppleWebKit/525 (KHTML, Someone, somewhere truncated this live user agent. Will the detection ecosystem pay the price? Detected As: Nokia E72 Device Match Failed Nokia E72 WURFL fails to find a similarity. fe64fb601209ea9c48e39bf5ebe9c c6f NokiaE72/021.024 (SymbianOS/9.3; U; Series60/3.2 Mozilla/5.0; Profile/MIDP-2.1 Configuration/CLDC-1.1) AppleWebKit/525 (KHTML, nokia_generic_series60 1557013 Nokia 60 NokiaE72 WURFL detects Series 60, but only version 1.0. Screensize: 320x240 176x208 320x240 Generic Series 60 screensize is the order of the day. MIDP version: 2.1 1.0 2.1 MP3 Support: Yes No Yes 3GPP Support: Yes No Yes Detected Browser: Series 60 3.2 Nokia Not Available 3.2 means Series 60 3 rd Edition Feature Pack 2. DetectRight detected this as the default series 60 3.2 browser, but Safari might have been more appropriate: or at last Nokia Apple Webkit. Detected OS: Symbian OS 9.3 Symbian OS Symbian OS v9.3 9.3 DR and DA got this right, though DA has the version number twice. Detected DP: Series 60 3.2 60 Series 60 3 There s a fair amount of difference between Series 60 3.0 and 3.2. Performance Mark: 4/5 1/5 3.5/5
Nokia 5320 XpressMusic(Series 60 3.2) Mozilla/5.0 (symbianos/9.3;u; Series60/3.2 Nokia5320d-1/04.13; Profile/MIDP-2.1 Configuration/CLDC-1.1) AppliWebKit/413(KHTML, like Gecko) Safari/413 You d think it would be simple. You re probably wrong. Note the misspelling of AppleWebKit, presumably caused by an engineer s typo @ Nokia. Detected As: Nokia 5320 XpressMusic Device Match Failed Nokia N97 mini This is eccentric from DA. 588ce77c0bf3e31272edd940588e a00b Mozilla/5.0 (symbianos/9.3;u; Series60/3.2 Nokia5320d-1/04.13; Profile/MIDP-2.1 Configuration/CLDC-1.1) AppliWebKit/413(KHTML, like Gecko) Safari/413 nokia_generic_series60 1666480 Back to Series 60 1.0 for DA. Nokia 60 Mozilla/5.0 (s So all Mozilla/5.0 (s user agents are the Nokia N97 Mini? That seems really random. Screensize: 240x320 176x208 360x640 DR got it. MIDP version: 2.1 1.0 2.1 MP3 Support: Yes No Yes 3GPP Support: Yes No No As an aside, I m sure the N97 Mini does support video. Detected Browser: Safari 413 Nokia Not Available Detected OS: Symbian OS 9.3 Symbian OS Symbian OS v9.4 9.4 I m sure Symbian 9.4 is a great OS. But it s not what we ve got here. Detected DP: Series 60 3.2 60 Series 60 5 Performance Mark: 5/5 1/5 2/5 (for Java and MP3) Odd.
Nokia N95 (Series 60 3.1) Mozilla/5.0 (Windows; U; Windows NT 5.1; en-us; rv:1.9.0.16) Gecko/2010010414 Mozilla/5.0 (SymbianOS/9.2; U; Series60/3.1 NokiaN95/10.0.010; Profile/MIDP-2.0 Configuration/CLDC-1.1 ) AppleWebKit/413 (KHTML, like Gecko) Safari/413 Flock/2.5.6 Possibly the longest user agent in the world, but still real. This UA really tests an engine s ability to adapt. Detected As: Nokia N95 Device Match Failed Unknown Vendor Unknown Model a71d0d62d9bbb01a7d611cb64 2e286a8 Mozilla/5.0 (Windows; U; Windows NT 5.1; en-us; rv:1.9.0.16) Gecko/2010010414 Mozilla/5.0 (SymbianOS/9.2; U; Series60/3.1 NokiaN95/10.0.010; Profile/MIDP-2.0 Configuration/CLDC-1.1 ) AppleWebKit/413 (KHTML, like Gecko) Safari/413 Flock/2.5.6 nokia_generic_series60 Nokia 60 Mozilla/5.0 (Windows; U; Windows NT 5.1; en-us; rv:1.9 Too rich for DA and WURFL s blood.. WURFL got closer, with a Series 60 match, though that s not as much help as it was. Screensize: 240x320 176x208 Not Available DR was the only one who managed to connect the UA to the right data. MIDP version: 2.0 1.0 No MP3 Support: Yes No No 3GPP Support: Yes No No Detected Browser: Safari 413 Nokia Not Available But what is Flock? Detected OS: Symbian OS 9.2 Symbian OS Not Available Detected DP: Series 60 3.1 60 Performance Mark: 5/5 1/5 0/5 Summary of detection performance (human generated)
Nokia N8-00 (Symbian ^3, Series 60 5.2) Mozilla 5.0 (Symbian/3; Series60/5.2 NokiaN8-00/00.0.000; Profile/MIDP-2.1 Configuration/CLDC-1.1) AppleWebKit/525 (KHTML, like Gecko) Version/3 BrowserNG/7.2.5.1 3gpp-gba A newie with Symbian s new flagship OS. Detected As: Nokia N8-00 Device Match Failed Nokia N8-00 Brand spanking new, but DeviceAtlas has it. 60fbf08dd90c2349dfb056c1dd06f aeb Mozilla 5.0 (Symbian/3; Series60/5.2 NokiaN8-00/00.0.000; Profile/MIDP-2.1 Configuration/CLDC-1.1) AppleWebKit/525 (KHTML, like Gecko) Version/3 BrowserNG/7.2.5.1 3gpp-gba nokia_generic_series60 1933520 Screensize: 360x640 176x208 360x640 MIDP version: 2.1 1.0 2.1 Nokia 60 Mozilla 5.0 (S Another odd string match. Is that really all DA used? MP3 Support: Yes No Yes 3GPP Support: Yes No No No video?? Come on, DA. Detected Browser: BrowserNG 7.2 Nokia Not Available What all the best Nokia devices are seen with these days Detected OS: OS Symbian^3 Symbian OS http://sw.nokia.com/dp- 1/OperatingSystem/symbian_os http://sw.nokia.com/dp- 1/OperatingSystem/symbian.os This must be the new Symbian^3 OS. DA s entry appears to be a cut/paste gone wrong, or an incorrectly programmed data retrieval system from Forum Nokia. Detected DP: Series 60 5.2 60 At least you can rely on Series 60 version numbering. Performance Mark: 5/5 1/5 3.5/5 DA is simply more up-to-date than WURFL with this, probably because updates are continuous rather than bulked.
Some Conclusions Detection misses have potentially expensive consequences, as does missing data. If a detection engine is only updated periodically, it places a great deal of importance on its deductive powers. Legacy recognition infrastructures are too simplistic to cope with the entirety of the useragent detection ecosystem. Organisations are advised to audit or assess their device detection systems for fitness of purpose.
Additional This document is released under Creative Commons Attribution-Noncommercial-No Derivative Works 2.0 The Author, Chris Abbott, is also the author of the DetectRight system, and has been working in mobile detection for five years. In this report, strict attention has been paid to fairness in judging device detection performance only on objective characteristics in the belief that objective data is the ultimate arbiter of quality.
Further information For an presentation of how device database deficiencies affect real-world performance for major brands, see Epic Fails for Major Brands For a presentation illustrating the differences between the major device databases, see Device Databases To compare DetectRight with other device detection engines in real-time, see the Detector Inspector