The sieved command is part of the Sieve implementation for the Dovecot secure
IMAP server. Sieve (RFC 5228) is a simple and highly extensible language for filtering
e-mail messages. It can be implemented for any type of mail access protocol, mail
architecture and operating system. The language cannot execute external programs and in
its basic form it does not provide the means to cause infinite loops, making it suitable
for running securely on mail servers where mail users have no permission run arbitrary programs.
Using the sieved command, Sieve binaries, which are produced for instance by
sievec(1), can be transformed into a human-readable textual representation. This can
provide valuable insight in how the Sieve script is executed. This is also particularly useful
to view corrupt binaries that can result from bugs in the Sieve implementation. This tool is
intended mainly for development purposes, so normally system administrators and users will not
need to use this tool.
The sieve-binary argument specifies the Sieve binary file that needs to be dumped. The
optional out-file argument specifies where the output must be written. If omitted, the
output is written to stdout.
The format of the output is not explained here in detail, but it should be relatively easy
to understand. The Sieve binaries comprise a set of data blocks, each of which can contain
arbitrary data. For the base language implementation two blocks are used: the first containing
a specification of all required language extensions and the second containing the main Sieve
program. Compiled Sieve programs are represented as flat byte code and therefore the dump of
the main program is a disassembly listing of the interpreter operations. Extensions can define
new operations and use additional blocks. Therefore, the output of sieved depends greatly
on the language extensions used when compiling the binary.