libfprint/doc/advanced-topics.xml
Bastien Nocera b3fe4a1e91 docs: Update API documentation
Fixes to layout, dead links, typography, and more.

Thanks to Benjamin Berg <bberg@redhat.com> for the thorough review
2018-05-18 05:51:58 +02:00

115 lines
5.3 KiB
XML

<?xml version="1.0"?>
<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.3//EN"
"http://www.oasis-open.org/docbook/xml/4.3/docbookx.dtd" [
]>
<chapter id="advanced-topics" xmlns:xi="http://www.w3.org/2003/XInclude">
<title>Advanced Topics</title>
<refsect2 id="compatibility">
<title>Device and print compatibility</title>
<para>
Moving off generic conceptual ideas and onto libfprint-specific
implementation details, here are some introductory notes regarding how
libfprint copes with compatibility of fingerprints.
</para>
<para>
libfprint deals with a whole variety of different fingerprint readers and
the design includes considerations of compatibility and interoperability
between multiple devices. Your application should also be prepared to
work with more than one type of fingerprint reader and should consider that
enrolled fingerprint X may not be compatible with the device the user has
plugged in today.
</para>
<para>
libfprint implements the principle that fingerprints from different devices
are not necessarily compatible. For example, different devices may see
significantly different areas of fingerprint surface, and comparing images
between the devices would be unreliable. Also, devices can stretch and
distort images in different ways.
</para>
<para>
libfprint also implements the principle that in some cases, fingerprints
<emphasis>are</emphasis> compatible between different devices. If you go and buy two
identical fingerprint readers, it seems logical that you should be able
to enroll on one and verify on another without problems.
</para>
<para>
libfprint takes a fairly simplistic approach to these issues. Internally,
fingerprint hardware is driven by individual drivers. libfprint enforces
that a fingerprint that came from a device backed by driver X is never
compared to a fingerprint that came from a device backed by driver Y.
</para>
<para>
Additionally, libfprint is designed for the situation where a single driver
may support a range of devices which differ in imaging or scanning
properties. For example, a driver may support two ranges of devices which
even though are programmed over the same interface, one device sees
substantially less of the finger flesh, therefore images from the two
device types should be incompatible despite being from the same driver. To
implement this, each driver assigns a <emphasis>device type</emphasis> to each device
that it detects based on its imaging characteristics. libfprint ensures that
two prints being compared have the same device type.
</para>
<para>
In summary, libfprint represents fingerprints in several internal structures
and each representation will offer you a way of determining the
<ulink url="#driver_id">driver ID</ulink> and <ulink url="#device-types">devtype</ulink> of the print in
question. Prints are only compatible if the driver ID <emphasis role="strong">and</emphasis> devtypes
match. libfprint does offer you some "is this print compatible?" helper
functions, so you don't have to worry about these details too much.
</para>
</refsect2>
<refsect2 id="driver_id">
<title>Driver IDs</title>
<para>
Each driver is assigned a unique ID by the project maintainer. These
assignments are
<ulink url="http://www.reactivated.net/fprint/Driver_ID_assignments">
documented on the wiki</ulink> and will never change.
</para>
<para>
The only reason you may be interested in retrieving the driver ID for a
driver is for the purpose of checking if some print data is compatible
with a device. libfprint uses the driver ID as one way of checking that
the print you are trying to verify is compatible with the device in
question - it ensures that enrollment data from one driver is never fed to
another. Note that libfprint does provide you with helper functions to
determine whether a print is compatible with a device, so under most
circumstances, you don't have to worry about driver IDs at all.
</para>
</refsect2>
<refsect2 id="device-types">
<title>Device types</title>
<para>
Internally, the <ulink url="libfprint-Driver-operations.html#libfprint-Driver-operations.description">driver</ulink> behind a device assigns a 32-bit
<emphasis>devtype</emphasis> identifier to the device. This cannot be used as a unique
ID for a specific device as many devices under the same range may share
the same devtype. The devtype may even be 0 in all cases.
</para>
<para>
The only reason you may be interested in retrieving the devtype for a
device is for the purpose of checking if some print data is compatible
with a device. libfprint uses the devtype as one way of checking that the
print you are verifying is compatible with the device in question - the
devtypes must be equal. This effectively allows drivers to support more
than one type of device where the data from each one is not compatible with
the other. Note that libfprint does provide you with helper functions to
determine whether a print is compatible with a device, so under most
circumstances, you don't have to worry about devtypes at all.
</para>
</refsect2>
</chapter>