[fprint] Question regarding fp_print_data_get_data()

[Geoffrey Teale]

On Jun 25, 2008, at 11:41 AM, Mehmet Ali Akmanalp wrote:
>

--- %< ---
>
> This is funny, I've been working on the exact same device. I guess  
> it's popular since it's cheaper than its counterparts or something.  
> Some tips: I've been working with the device out of the laptop  
> enclosure, with a usb cable soldered on. It seems to be harder to  
> make it enroll or verify properly when it's out of the enclosure,  
> since the plastic curved dent that guides your finger is not there.  
> It's a bit more consistent otherwise. Also, people with relatively  
> drier skin like me have a harder time getting the reader to detect  
> their finger properly. The hackish solution: lick your finger!
>
> ~ Mehmet Ali Akmanalp
>


Yes - it does seem hideously popular - it's hard to find a device  
suitable for my purposes that doesn't feature it.  Talking to OEMs it  
seems that price and size and the features they desire, plus it does  
seem to work a little better under Windows with the Authentec  
proprietary code.

  I haven't tried it outside an enclosure.   I have used the device on  
the Motion Computing C5 (damn near impossible as it is mounted on the  
thing edge of the device) and in the Data Ltd, Inc, DLI8300 (much  
better).   I am able to get good scans every time (with practice), but  
end users have reported a great deal of difficulty in using it.   
Particularly it is very easy to enroll a fingerprint that you will  
never be able to verify against.   I can also make libfprint segfault  
on demand by swiping my finger very quickly, or dabbing it on the  
device.  My patch (in a previous mail) fixed this segfault and added  
additional error codes for wiping to fast or too slow.

I have an additional issue that people may overlook when using this  
for personal reasons -  I have to use this device to provide  
authenticated access to data capture system for clinical trials, that  
means this device has to be really secure and well tested.   WIth the  
stock libfprint code I have (exactly once) achieved a false positive.   
I have to make the scanner more picky about what it calls a good scan  
and work to make simultaneously more secure and easier to use.    An  
unenviable task I am sure you will agree!

-- 
Geoffrey Teale
Software and Technology Consultant, München
tealeg at member.fsf.org