camlib_devel.sgml

<!doctype linuxdoc system>

<!-- LyX 1.1 created this file. For more info see http://www.lyx.org/ -->
<article>
<title>
gPhoto2 Camera Library Developer's Guide
</title>
<author>
Scott Fritzinger
2000-07-26
Revision 1
</author>
<toc>
<sect>
Reverse Engineering the Camera Protocol
<p>
The most difficult part for most developers is obtaining the transfer protocol.
 If the OEM's are lucky enough, they will simply provide us with the protocol
 specifications for their cameras and the drivers will be written at no cost
 to them. Most OEM's refuse to do so though, citing trade secrets or company
 policy; this is truly unfortunate in that they have effectively told their
 own customers who use operating systems other than Windows and the Mac that
 they don't want their future business and that they aren't valued customer
 to begin with.
</p>
<p>
When OEM's do not cooperate, the developer is left to determine the protocol
 him/herself through reverse engineering. 
</p>
<sect1>
Sniffing the Protocol
<p>
What follows are the most common setups for sniffing camera protocol traffic.
 In all setups, a host computer runs the native camera drivers.; typically,
 the Windows serial port drivers are used for reverse engineering. The drivers
 are run through a series of functions that include getting a picture index,
 downloading thumbnails, download full images, deleting images, camera configuration
 options, in addition to any other features a camera might have. During these
 operations, one or more of the following methods are used to capture the communication
 between the host computer and the camera.
</p>
<sect2>
Serial Repeater
<p>
A serial repeater consists of the host computer, a computer used as a repeater,
 and the camera. The setup is shown in figure .
</p>
<p>
The repeater runs special software which reads data from one serial port,
 logs the communication, and then outputs the data to the other serial port.
 Data that is from the host computer to the camera and from the camera to the
 host computer is logged sequentially in a single log file. Information logged
 includes hexadecimal data values, direction of the communication, as well as
 time stamps for synchronization. An example sniffer to use for this configuration
 is "sersniff".
</p>
<sect2>
"Y" Serial Cable
<p>
To avoid using two computers, a Y serial cable can be used. The "trunk" end
 of the serial cable attaches to the camera's serial transfer cable, while the
 two "branches" plug in to two serial ports on the host computer. Figure shows
 this setup.
</p>
<p>
The camera drivers use one of the serial ports on the host computer, while
 the other port is opened with a hexadecimal monitor application that dumps
 all communications on the port to a file. The downside to this approach is
 the developer would have to determine which sets of data was generated by the
 camera or the host computer. Also, a Y cable would have to be either built
 or purchased from an electronics supply store.
</p>
<sect2>
Virtual Device Driver Hooks
<p>
The Windows platform allows virtual device drivers to "hook" into other drivers
 to provide additional functionality or feature enhancements. A combination
 GUI and device driver named PortMon by Systems Internals is a communications
 debugging utility that hooks into the existing Windows serial device driver
 (vcomm.vxd) and logs communications. Figure shows this equipment arrangement.
</p>
<p>
This setup allows the developer to not use any extra hardware by simply
 relying on software. This is perhaps the easiest method for capturing camera
 data.
</p>
<sect1>
Making Sense Out of the Protocol
<p>
What follows are some pointers on decoding camera protocols. It uses a
 protocol that isn't really any camera protocol in particular, but should demonstrate
 some commonalities between most camera protocols.
</p>
<p>
<enum>
 <item>
Cameras like to ping. This is the in the form of an &quot;ACK&quot;&quot;
 command that is different for different cameras. Basically, it is usually a
 short packet (probably 1 byte) that is sent both ways in order for the camera
 to know the computer is there or vice versa. It is also sometimes used to wake
 up a camera that has gone into power-save mode. It usually starts out the communications,
 as well as confirms each packet in any sort of &quot;mass&quot; transfer. The
 opposite, a &quot;NAK&quot;, is sent to basically say the last packet was not
 received, or an error has occurred. Again, this is usually just a single byte
 as well.

Example:
Computer: 01
Camera : 01

The Camera sent an ACK (&quot;01&quot;)
 and the Computer responded with an ACK as well.
 <item>
Transfers are usually in &quot;reverse network order&quot;, meaning least
 significant bytes come before most significant bytes. For example, "00 08" should
 actually be reassembled as "08 00". 
 <item>
Most protocols use starting and stopping bytes. 

Example:
Computer: 03
 50 00 0f e0 04 
Camera : 03 03 00 3f 03 04 
Computer: 01 

For this example,
 notice the packets begin with &quot;03&quot; and end with &quot;04&quot; (don't
 pay attention to what is between them). Also notice the Computer sent an &quot;ACK&quot;
 to confirm it got the packet.
 <item>
Packets usually have a &quot;command&quot; byte, which tells either the
 computer or the camera what to do. Let's say you told the software to retrieve
 the number of pictures, which at the time happened to be &quot;8&quot;, and
 you got the following:

Computer: 03 01 00 00 00 04 
Camera : 03 01 00 00 08
 04 
Computer: 01

In this example, you notice the &quot;03&quot; and &quot;04&quot;
 specifying the start and stop of the packet. Also, you notice the second byte
 in the Computer packet is &quot;01&quot;. The camera responds with the above
 packet, and low and behold, you see the number 8 in the same packet. It would
 appear, initially, that the second byte is used as a command byte, and that
 &quot;01&quot; specifies the camera to return the number of pictures. This
 may very well be right, but don't jump into it yet. Make sure you look at a
 bunch of similar situations to confirm this. (Again, notice the &quot;ACK&quot;
 sent by the computer).
 <item>
Most protocols have a &quot;data size&quot; byte(s) in data packets. Let's
 say that you told the camera to retrieve thumbnail 8 and you get the following:

Computer:
 03 02 00 00 08 04
Camera : 03 02 00 0F (15 bytes) 04
Computer: 01 

OK, here's
 a brief breakdown of this transaction:

-Looks like the command to retrieve
 a thumbnail is &quot;02&quot; (2nd byte in the computer packet), and that the
 byte that is &quot;08&quot; specifies which thumbnail to return.
-The camera
 responds with a &quot;02&quot; in the command field, specifying it is returning
 a thumbnail, and then sends &quot;0F&quot;, and 15 bytes of data. 
-It looks
 like the byte &quot;0F&quot; specifies how many bytes are after it in the same
 packet. This is a data size byte. 
(Note: this is a simplistic example. No
 thumbnail will only be 15 bytes :) this leads up to the next thing to consider)
 <item>
Most protocols have an &quot;order&quot; or &quot;counter&quot; byte. This
 is used so that, in large data transfers where the picture may be split up
 into several different packets, the computer knows how to reassemble all the
 data. The entire thumbnail more than likely will not be contained in a single
 packet for logistical reasons, so they break up the data into many different
 packets and give each packet a unique number (or "order" byte). Let's say you
 told your camera to return thumbnail 8 (which is, as mentioned, pretty big),
 and you get the following:

Computer: 02 03 00 00 08 03 
Camera : 02 03 00
 0F (15 bytes) 03 
Computer: 01
Camera : 02 03 01 0F (15 bytes) 03 
Computer:
 01
Camera : 02 03 02 0F (15 bytes) 03
Computer: 01
... 5 more packets and ACKs
 ...
Camera : 02 03 08 09 (9 bytes) 03
Computer: 01

You notice that the 3rd
 byte of each of the camera packets increments with each packet sent from the
 camera. This looks like it is an order (counter) byte. the computer can then
 reassemble the data from all the packets in order to reproduce the image.
 <item>
Most protocols have some sort of error detection byte(s) at the end of
 the packet. This is usually a simple checksum (summation of bytes), or a CRC
 (a somewhat complex algorithm that reduces the probability of mis-diagnosing
 a packet with errors by magnitudes). These bytes can take into account only
 the data, or maybe the entire packet excluding those error detection bytes.
 If this isn't a known scheme, this winds up being the hardest part of reimplementing
 the protocol. Lets take the above example again, this time we'll add a couple
 bytes on the end for error detection:

Computer: 02 03 00 00 08 03
Camera :
 02 03 00 0F (15 bytes) 0f 02 03
Computer: 01
Camera : 02 03 01 0F (15 bytes)
 0e 00 03
Computer: 02
Camera : 02 03 01 0F (15 bytes) fa d0 03
Computer: 01
Camera
 : 02 03 02 0F (15 bytes) fa d0 03
Computer: 01
... 5 more packets and ACKs
Camera
 : 02 03 08 09 (9 bytes) d7 38 03
Computer: 01

Notice how the error detection
 bytes are usually different for each packet. These may be checksums, or CRC's,
 or something else. Only way to find out really is to try each one, on different
 combinations of packet parts (data, order byte, command byte, etc...) and see
 if you get the same thing. Try this on the shorter packets to make life easier.

Look
 at one more thing that sticks out in this transaction: for packet with order
 byte "01", the Computer responded with a &quot;02". and the Camera then resent
 the same packet it just did. This shows that the NAK byte is &quot;02&quot;,
 and this could happen because maybe the error detection bytes didn't match
 with the data, or maybe something else happened. either way, the camera resent
 the last packet, and now you know how the camera can recover from transfer
 errors. If you didn't get the packet you were expecting, send the camera a
 NAK and it will resend the same packet again.
</enum>
</p><sect>
Understanding the gPhoto2 Design
<p>
The gPhoto2 design is the same three-tiered structure that has worked extremely
 well in the past with other software packages. Here is a listing of the 3 tiers:
</p>
<p>
<itemize>
 <item>
the camera library
 <item>
the I/O library
 <item>
the front-end
 <item>
the "core"
</itemize>
</p><sect1>
Role of the Camera Library
<p>
The camera library is in charge of talking directly with the camera. The
 library uses the gPhoto2 Camera API in order to provide a common access-method
 for the library itself. Being dynamically linked, the libraries are loaded
 at run-time depending on the camera model the end-user would like to access.
 
</p>
<p>
In order to provide flexilibity with variations in camera design, there
 are camera "abilities" which list, well, the abilities of each camera model.
 Some camera may support serial port connections only, while others may be able
 to use USB and a serial port. We've run into cameras that don't support thumbnailing
 on the camera so there is an "abilities" field to specify whether or not the
 camera supports thumbnailing. The "abilities" also list other things such as
 supported serial transfer speeds, file deletion, and other functionality.
</p>
<p>
The camera libraries only make functions calls to the I/O library and to
 the gPhoto2 core.
</p>
<p>
There is more information on the specifics of the camera library in section
 3 of this document.
</p>
<sect1>
Role of the I/O Library
<p>
The gPhoto2 I/O library is a platform-independent communications library
 that support serial, parallel, USB, firewire, and network connections. It is
 a work-in-progress with a constantly expanding list of supported platforms.
 This library uses the gPhoto2 I/O library API for accessing communications
 devices. It enumerates the devices available on a system, and provides read/write
 access.
</p>
<p>
The camera libraries all use the I/O library for communications with the
 cameras. By doing having all communications go through a single library, the
 camera libraries become as portable as the I/O library. Porting gPhoto2 to
 other platforms become extremely easy.
</p>
<p>
There is more information on the specifics of the I/O library in section
 3 of this document.
</p>
<sect1>
Role of the Front-end
<p>
The front-end is the application that the user interacts with. It is usually
 a command-line program, or a graphical point-and-click interface. The front-end
 talks only with the gPhoto2 core in order to retrieve pictures and perform
 other functions with the camera.
</p>
<sect1>
Role of the gPhoto2 Core
<p>
The gPhoto2 "core" is the heart of gPhoto2. It provides services to both
 the camera libraries and the front-ends. Most of the services deal with error-checking
 and enumeration of devices (cameras, I/O devices, etc...). The core performs
 validity checking on data passed to/from the front-end or the camera library.
</p>
<p>
You could consider the core a translator/interpreter/spell-checker/army-general
 in the "big picture" of gPhoto2. It does the grunt-work and performs the coordination
 of the other parts.
</p>
<sect>
Implementing the Library
<p>
gPhoto2 camera libraries use the gPhoto2 Camera API (CAPI) for implementation.
 Here is a listing of the CAPI functions: 
</p>
<p>
camera_id 
</p>
<p>
camera_abilities 
</p>
<p>
camera_init 
</p>
<p>
camera_exit
</p>
<p>
camera_folder_list
</p>
<p>
camera_file_list 
</p>
<p>
camera_file_get
</p>
<p>
camera_file_get_preview
</p>
<p>
camera_file_put 
</p>
<p>
camera_file_delete 
</p>
<p>
camera_config_get
</p>
<p>
camera_config_set
</p>
<p>
camera_capture
</p>
<p>
camera_summary
</p>
<p>
camera_manual
</p>
<p>
camera_about

</p>
<p>
Section 3.1 details the purpose of each of these functions, while Section
 3.2 discusses how to use the I/O library.
</p>
<sect1>
Camera API
<p>
The CAPI provides the full set of functions for doing various tasks with
 the camera. All CAPI functions return either GP_OK for successful execution
 , or GP_ERROR for a failure of execution
</p>
<p>
What follows is a listing of the functions, including prototypes and data
 exchange:
</p>
<sect2>
camera_id
<p>
<bf>Purpose: </bf>Retrieve the unique id for the camera library.
</p>
<p>
<bf>Prototype: </bf>int camera_id (CameraText *id); 
</p>
<p>
<bf>Arguments: </bf>
</p>
<p>
CameraText *id : unique string to represent the camera library

</p>
<p>
In order to guarantee that only once instance of the camera library is
 loaded for each instance of the core, the camera library must copy a unique
 string into the "id". Please consult the gPhoto developers to determine which
 string you should use.

</p>
<p>
<bf>Example:</bf>
</p>
<p>
int camera_id(CameraText *id) &lcub;
</p>
<p>
strcpy(id-&gt;text, "my-unique-string");
</p>
<p>
return (GP_OK);
</p>
<p>
&rcub;
</p>
<sect2>
camera_abilities
<p>
<bf>Purpose: </bf>Retrieve the list of supported cameras and the abilities for each
 camera
</p>
<p>
<bf>Prototype: </bf>int camera_abilities (CameraAbilitiesList *list); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<p>
CameraAbilities *abilities : the list of abilities for the supported cameras
</p>
<p>
int *count : the number of 
</p>
<sect2>
camera_init
<p>
<bf>Purpose: </bf>Initialize the camera
</p>
<p>
<bf>Prototype: </bf>int camera_init (Camera *camera, CameraInit *init); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_exit 
<p>
<bf>Purpose: </bf>Close the camera
</p>
<p>
<bf>Prototype: </bf>int camera_exit (Camera *camera); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_file_list
<p>
<bf>Purpose: </bf>List the files in a particular folder on the camera
</p>
<p>
<bf>Prototype: </bf>int camera_file_list(Camera *camera, CameraList *list, char
 *folder); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_folder_list
<p>
<bf>Purpose: </bf>List the subfolders in a particular folder on the camera
</p>
<p>
<bf>Prototype: </bf>int camera_folder_list(Camera *camera, CameraList *list, char
 *folder); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_file_get
<p>
<bf>Purpose: </bf>Retrieve a file from the camera
</p>
<p>
<bf>Prototype: </bf>int camera_file_get (Camera *camera, CameraFile *file, char
 *folder, char *filename); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_file_get_preview
<p>
<bf>Purpose: </bf>Retrieve a file's preview from the camera
</p>
<p>
<bf>Prototype: </bf>int camera_file_get_preview (Camera *camera, CameraFile *file,
 char *folder, char *filename); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_file_put
<p>
<bf>Purpose: </bf>Place (upload) a file to the camera
</p>
<p>
<bf>Prototype: </bf>int camera_file_put (Camera *camera, CameraFile *file, char
 *folder); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_file_delete
<p>
<bf>Purpose: </bf>Delete a file from the camera
</p>
<p>
<bf>Prototype: </bf>int camera_file_delete (Camera *camera, char *folder, char *filename);
 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_config_get
<p>
<bf>Purpose: </bf>Retrieve the configuration window.
</p>
<p>
<bf>Prototype: </bf>int camera_config_get (Camera *camera, CameraWidget *window);
 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_config_set
<p>
<bf>Purpose: </bf>Set camera configuration
</p>
<p>
<bf>Prototype: </bf>int camera_config_set (Camera *camera, CameraSetting *setting,
 int count); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_capture
<p>
<bf>Purpose: </bf>Retrieve live data from the camera
</p>
<p>
<bf>Prototype: </bf>int camera_capture (Camera *camera, CameraFile *file, CameraCaptureInfo
 *info); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_summary
<p>
<bf>Purpose: </bf>Retrieve the camera summary information
</p>
<p>
<bf>Prototype: </bf>int camera_summary (Camera *camera, CameraText *summary); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_manual
<p>
<bf>Purpose: </bf>Retrieve the camera user's guide (manual)
</p>
<p>
<bf>Prototype: </bf>int camera_manual (Camera *camera, CameraText *manual); 
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect2>
camera_about
<p>
<bf>Purpose: </bf>Retrieve information about the camera library
</p>
<p>
<bf>Prototype: </bf>int camera_about (Camera *camera, CameraText *about);
</p>
<p>
<bf>Arguments: </bf>d
</p>
<sect1>
The gPhoto2 I/O Library


</article>