\input texinfo                      @c -*-texinfo-*-
@c %**start of header
@setfilename assuan.info

@macro copyrightnotice
Copyright @copyright{} 2002, 2003 Free Software Foundation, Inc.
@end macro
@macro permissionnotice
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version. The text of the license can be found in the
section entitled ``Copying''.
@end macro

@include version.texi

@settitle Developing with Assuan

@c Create a separate index for command line options.
@defcodeindex op
@c Merge the standard indexes into a single one.
@syncodeindex fn cp
@syncodeindex vr cp
@syncodeindex ky cp
@syncodeindex pg cp
@syncodeindex tp cp

@c printing stuff taken from gcc.
@macro gnupgtabopt{body}
@code{\body\}
@end macro
@macro gnupgoptlist{body}
@smallexample
\body\
@end smallexample
@end macro
@c Makeinfo handles the above macro OK, TeX needs manual line breaks;
@c they get lost at some point in handling the macro.  But if @macro is
@c used here rather than @alias, it produces double line breaks.
@iftex
@alias gol = *
@end iftex
@ifnottex
@macro gol
@end macro
@end ifnottex


@c Change the font used for @def... commands, since the default
@c proportional one used is bad for names starting __.
@tex
\global\setfont\defbf\ttbshape{10}{\magstep1}
@end tex

@c %**end of header

@ifnottex
@dircategory GNU Libraries
@direntry
* Assuan: (assuan).        An IPC library for non-persistent servers.
@end direntry
This file documents the use and the internals of Assuan.

This is Edition @value{EDITION}, last updated @value{UPDATED}, of
@cite{The `Developing with Assuan' Manual}, for Version @value{VERSION}.
@sp 1
Published by the Free Software Foundation@*
59 Temple Place - Suite 330@*
Boston, MA 02111-1307 USA
@sp 1
@copyrightnotice{}
@sp 1
@permissionnotice{}
@end ifnottex

@setchapternewpage odd

@titlepage
@title Developing with Assuan
@subtitle Version @value{VERSION}
@subtitle @value{UPDATED}
@author Werner Koch @code{(wk@@gnupg.org)}

@page
@vskip 0pt plus 1filll
@copyrightnotice{}
@sp 2
@permissionnotice{}
@end titlepage
@summarycontents
@contents
@page


@ifnottex
@node Top
@top Introduction
@cindex introduction

This manual documents how to exploit the Assuan library, a simple
interprocess communcation library.
@end ifnottex

@menu
* Introduction::        An introduction to and the motivation behind Assuan.
* Assuan::              Description of the Assuan protocol.
* Implementation::      Overview of the implementation.

Miscellaneous

* Library Copying::     GNU Lesser General Public License says
                        how you can copy and share Assuan.
* Copying::             How you can copy and share this manual.

Indices

* Option Index::        Index to command line options.
* Index::	        Index of concepts and symbol names.

@end menu



@node Introduction
@chapter Introduction to Assuan

In an ideal world, Assuan is irrelevant.  Assuan's primary use is to
allow a client to interact with a non-persistent server.  Using
Assuan, this is accomplished by forking a subprocess and communicating
with it via, for example, a pipe or unix domain socket.  This method
is neither elegant nor efficient especially when there is a lot of
data spread across several transactions: not only is there a penalty
for an increased number of context switches, but also a significant
amount of data is @var{memcpy}ed from the client to a file descriptor
and from the file descriptor to the server.  Despite these and other
disadvantages, this type of client/server communication can be useful:
the client is completely separate from the server; they are in
different address spaces.  This is especially important in situations
where the server must have a known degree of reliability and data must
be protected: as the Assuan protocol is well defined and clients
cannot corrupt the servers' address space, auditing become much
easier.

Assuan was developed for use by the GNU Privacy Guard, GnuPG, to
prevent potentially buggy clients from unwittingly corrupting
sensitive transactions or compromising data such as a secret key.
Assuan permits the servers, which do the actual work, e.g. encryption
and decryption of data using a secret key, to be developed
independently of the user interfaces, e.g. mail clients and other
encryption front ends.  Like a shared library, the interface is well
defined and any number of front ends can use it; however, unlike a
shared library, the client cannot see or touch the server's data.  As
with any modular system, Assuan helps keep the servers small and
understandable help to make code more understandable and less error
prone.

Assuan is not, however, limited to use with GnuPG servers and clients:
it was design to be flexible enough to meet the demands of almost any
transaction based environment with non-persistent servers.

@node Assuan
@chapter Description of the Assuan protocol.

The architecture of the modular GnuPG system is based on several
highly specialized modules which compose a network of client/server
communication.  A common framework for intermodule communication is
therefore needed and should be implemented in a library.

Goals:

@itemize @bullet
@item Common framework for module communication
@item Easy debugging
@item Easy module testing
@item Extendible
@item Optional authentication and encryption facility
@item Usable by access external hardware
@end itemize


Design criteria:

@itemize @bullet
@item Client server with back channel
@item Use a mainly text based protocol
@item Escape certain control characters
@item Allow indefinite data length
@item Request confidentiality for parts of the communication
@item Dummy module to allow direct linking of client and server
@item Inline data or descriptor passing for bulk data
@item No protection against DoS needed
@item Subliminal channels are not an issue
@end itemize

@node Implementation
@chapter Implementation

The implementation is line based with a maximum line size of 1000
octects.  The default IPC mechanism are Unix Domain Sockets.

On a connect request the server responds either with an okay or an error
status.  For authentication check the server may send an Inquiry
Response prior to the first Okay, it may also issue Status messages.
The server must check that the client is allowed to connect, this is
done by requesting the credentials for the peer and comparing them to
those of the server.  This avoids attacks based on wrong socket
permissions.

It may choose to delay the first response in case of an error.  The
server never closes the connection - however the lower protocol may do
so after some time of inactivity or when the connection is in an error
state.

All textual messages are assumed to be in UTF-8 unless otherwise noted.


@menu
* Server responses::    Description of server responses.
* Client requests::     Description of client requests.
* Error codes::         List of error and status codes.
@end menu


@node Server responses
@section Server responses

@table @code
@item OK  [<arbitary debugging information>]
Request was successful.

@item ERR @var{errorcode} [<human readable error description>]
Request could not be fulfilled.  The error codes are mostly application
specific except for a few common ones.

@item S @var{keyword} <status information depending on keyword>
Informational output by the server, still processing the request.

@item # <string>
Comment line issued only for debugging purposes.  Totally ignored.

@item D <raw data>
Raw data returned to client. There must be exactly one space after the
'D'.  The values for '%', CR and LF must be percent escaped; this is
encoded as %25, %0D and %0A.  Only uppercase letters should be used in
the hexadecimal representation.  Other characters may be percent escaped
for easier debugging.  All these Data lines are considered one data
stream up to the OK or ERR response.  Status and Inquiry Responses
may be mixed with the Data lines.

@item INQUIRE @var{keyword}> <parameters>
Server needs further information from the client.  The client should
answer with a command which is allowed after an inquiry.  Note that the
server does not confirm that client command but either continues
processing or ends processing with an error status.  Not all commands
are allowed.
@end table


A client should only check the first letter of each line and then skip
over to the next token (except for data lines where the raw data starts
exactly after 2 bytes).  Lines larger than 1000 bytes should be
treated as a communication error. (The rationale for having a line
length limit is to allow for easier multiplexing of multiple channels).


@node Client requests
@section Client requests

The server waits for client requests after he sent an Okay or Error.
The client should not issue a request in other cases.

@example
@var{command} <parameters>
@end example

@var{command} is a one word string without preceding white space.
Parameters are command specific, CR, LF and the percent signs should be
percent escaped as described above.  To send a backslash as the last
character it should also be percent escaped.  Percent escaping is
allowed anywhere in the parameters but not in the command.  The line
ends with a CR, LF or just a LF.

Not yet implemented feature: If there is a need for a parameter list
longer than the line length limit (1000 characters including command and
CR, LF), the last character of the line (right before the CR/LF or LF)
must be a non-escape encoded backslash. The following line is then
expected to be a continuation of the line with the backslash replaced by
a blank and the line ending removed.

@example
D <raw data>
@end example

Raw data to the server. There must be exactly one space after the 'D'.
The values for '%', CR and LF must be percent escaped; this is encoded
as %25, %0D and %0A.  Only uppercase letters should be used in the
hexadecimal representation.  Other characters may be percent escaped for
easier debugging.  All these Data lines are considered one data stream
up to the OKAY or ERROR response.  Status and Inquiry Responses may be
mixed with the Data lines.

@example
END
@end example



Lines beginning with a @code{#} or empty lines are ignored.  This is
useful to comment test scripts.


Although the commands are application specific, some of them are used by
all protocols and partly directly supported by the Assuan library:

@table @code
@item CANCEL
This command is used for future extenxions.  It may today be used to
cancel outstanding requests in an asynchronous protocol.

@item BYE
Close the connect, the server will reply with an @code{OK}.

@item AUTH
Not yet specified as we don't implement it in the first phase.  See my
mail to gpa-dev on 2001-10-25 about the rationale for measurements
against local attacks.

@item RESET
Reset the connection but not any existing authentication.  The server
should release all resources associated with the connection.

@item END
Used by a client to mark the end of raw data.  The server may send END
to indicate a partial end of data.

@item HELP
Reserved for future extensions.

@item QUIT
Reserved for future extensions.

@end table


@node Error codes
@section Error codes

Here we keep a list of error codes used in any Assuan based
protocol.  The format is the string @code{ERR}, white space, the error
number, white space, a textual description of the error.

General error codes pertaining to the actual Assuan operations:

@table @code
@item 0  Success
@item 1  General error
@item 2  Out of core
@item 3  Invalid value
@item 4  Timeout
@item 5  Read error
@item 6  Write error
@item 7  Problem starting server
@item 8  Not a server
@item 9  Not a client
@item 10 Nested commands
@item 11 Invalid response
@item 12 No data callback
@item 13 No inquire callback
@item 14 Connect failed
@item 15 Accept failed
@end table

Error codes used as status codes in the Assuan protocol:

@table @code
@item 100 Not implemented
@item 101 Server fault (catch all error code)
@item 102 Invalid command
@item 103 Unknown command
@item 104 Syntax error
@item 105 Parameter error
@item 106 Parameter conflict
@item 107 Line too long
@item 108 Line not terminated
@item 109 No input
@item 110 No output
@item 111 Canceled
@item 112 Unsupported algorithm
@item 113 Server resource problem
@item 114 Server I/O error
@item 115 Server bug
@item 116 No data available
@item 117 Invalid data
@item 118 Unexpected command
@item 119 Too much data
@item 120 Inquire unknown
@item 121 Inquire error
@item 122 Invalid option
@item 123 Invalid index
@item 124 Unexpected status
@item 125 Unexpected data
@item 126 Invalid status
@item 128 Not confirmed
@end table

For historical reasons a few more error codes are defined in
@file{assuan.h}; they should not be used be new applications.

Errror codes in the range @var{ASSUAN_USER_ERROR_FIRST} to
@var{ASSUAN_USER_ERROR_LAST} may be used at the applications own
discretion. Error codes greater than 65535 are not defined by Assuan
and may also be used by applications --- note that the GnuPG system
maps libgpg-error codes into this range.



@include lgpl.texi
@include gpl.texi


@c ---------------------------------------------------------------------
@c Indexes
@c ---------------------------------------------------------------------

@node Option Index
@unnumbered Option Index

@printindex op

@node Index
@unnumbered Index

@printindex cp

@c ---------------------------------------------------------------------
@c Epilogue
@c ---------------------------------------------------------------------

@bye