tar/doc/sparse.texi

@c This is part of the paxutils manual.
@c Copyright (C) 2006 Free Software Foundation, Inc.
@c This file is distributed under GFDL 1.1 or any later version
@c published by the Free Software Foundation.

The notion of sparse file, and the ways of handling it from the point
of view of @GNUTAR{} user have been described in detail in
@ref{sparse}.  This chapter describes the internal format @GNUTAR{}
uses to store such files.

The support for sparse files in @GNUTAR{} has a long history.  The
earliest version featuring this support that I was able to find was 1.09,
released in November, 1990.  The format introduced back then is called
@dfn{old GNU} sparse format and in spite of the fact that its design
contained many flaws, it was the only format @GNUTAR{} supported 
until version 1.14 (May, 2004), which introduced initial support for
sparse archives in @acronym{PAX} archives (@pxref{posix}).  This
format was not free from design flows, either and it was subsequently
improved in versions 1.15.2 (November, 2005) and 1.15.92 (June,
2006). 

In addition to GNU sparse format, @GNUTAR{} is able to read and
extract sparse files archived by @command{star}.

The following subsections describe each format in detail.

@menu
* Old GNU Format::
* PAX 0::                PAX Format, Versions 0.0 and 0.1
* PAX 1::                PAX Format, Version 1.0
@end menu

@node Old GNU Format
@appendixsubsec Old GNU Format

The format introduced some time around 1990 (v. 1.09).  It was
designed on top of standard @code{ustar} headers in such an
unfortunate way that some of its fields overwrote fields required by
POSIX.

An old GNU sparse header is designated by type @samp{S}
(@code{GNUTYPE_SPARSE}) and has the following layout:

@multitable @columnfractions 0.10 0.10 0.20 0.20 0.40
@headitem Offset @tab Size @tab Name   @tab Data type   @tab Contents
@item          0 @tab 345  @tab        @tab N/A         @tab Not used.
@item        345 @tab  12  @tab atime  @tab Number      @tab @code{atime} of the file.
@item        357 @tab  12  @tab ctime  @tab Number      @tab @code{ctime} of the file .
@item        369 @tab  12  @tab offset @tab Number      @tab For
multivolume archives: the offset of the start of this volume.
@item        381 @tab   4  @tab        @tab N/A         @tab Not used.
@item        385 @tab   1  @tab        @tab N/A         @tab Not used.
@item        386 @tab  96  @tab sp     @tab @code{sparse_header} @tab (4 entries) File map.
@item        482 @tab   1  @tab isextended @tab Bool        @tab @code{1} if an
extension sparse header follows, @code{0} otherwise.
@item        483 @tab  12  @tab realsize @tab Number      @tab Real size of the file.
@end multitable

Each of @code{sparse_header} object at offset 386 describes a single
data chunk. It has the following structure: 

@multitable @columnfractions 0.10 0.10 0.20 0.60
@headitem Offset @tab Size @tab Data type   @tab Contents
@item          0 @tab   12 @tab Number      @tab Offset of the
beginning of the chunk.
@item         12 @tab   12 @tab Number      @tab Size of the chunk.
@end multitable

If the member contains more than four chunks, the @code{isextended}
field of the header has the value @code{1} and the main header is
followed by one or more @dfn{extension headers}.  Each such header has
the following structure:

@multitable @columnfractions 0.10 0.10 0.20 0.20 0.40
@headitem Offset @tab Size @tab Name   @tab Data type   @tab Contents
@item          0 @tab   21 @tab sp     @tab @code{sparse_header} @tab
(21 entires) File map.
@item        504 @tab    1 @tab isextended @tab Bool    @tab @code{1} if an
extension sparse header follows, or @code{0} otherwise.
@end multitable

A header with @code{isextended=0} ends the map.

@node PAX 0
@appendixsubsec PAX Format, Versions 0.0 and 0.1
@UNREVISED{}

There are two formats available in this branch.  The version @code{0.0}
is the initial version of sparse format used by @command{tar}
versions 1.14--1.15.1.  The sparse file map is kept in extended
(@code{x}) PAX header variables:

@table @code
@item GNU.sparse.size
Real size of the stored file

@item GNU.sparse.numblocks
Number of blocks in the sparse map

@item GNU.sparse.offset
Offset of the data block

@item GNU.sparse.numbytes
Size of the data block
@end table

The latter two variables repeat for each data block, so the overall
structure is like this:

@smallexample
@group
GNU.sparse.size=@var{size}      
GNU.sparse.numblocks=@var{numblocks} 
repeat @var{numblocks} times
  GNU.sparse.offset=@var{offset}    
  GNU.sparse.numbytes=@var{numbytes}  
end repeat
@end group
@end smallexample

This format presented the following two problems:

@enumerate 1
@item
Whereas the POSIX specification allows a variable to appear multiple
times in a header, it requires that only the last occurrence be
meaningful.  Thus, multiple ocurrences of @code{GNU.sparse.offset} and
@code{GNU.sparse.numbytes} are conficting with the POSIX specs.

@item
Attempting to extract such archives using a third-party @command{tar}s
results in extraction of sparse files in @emph{compressed form}.  If
the @command{tar} implementation in question does not support POSIX
format, it will also extract a file containing extension header
attributes.  This file can be used to expand the file to its original
state.  However, posix-aware @command{tar}s will usually ignore the
unknown variables, which makes restoring the file much more
difficult@FIXME-xref{how to extract sparse file using third-party @command{tar}s}.
@end enumerate

@GNUTAR{} 1.15.2 introduced sparse format version @code{0.1}, which
attempted to solve these problems.  As its predecessor, this format
stores sparse map in the extended POSIX header.  It retains
@code{GNU.sparse.size} and @code{GNU.sparse.numblocks} variables, but
instead of @code{GNU.sparse.offset}/@code{GNU.sparse.numbytes} pairs
it uses a single variable:

@table @code
@item GNU.sparse.map
Map of non-null data chunks.  It is a string consisting of
comma-separated values "@var{offset},@var{size}[,@var{offset-1},@var{size-1}...]" 
@end table

To address the 2nd problem, the @code{name} field in @code{ustar}
is replaced with a special name, constructed using the following pattern:

@smallexample
%d/GNUSparseFile.%p/%f
@end smallexample

The real name of the sparse file is stored in the variable
@code{GNU.sparse.name}.  Thus, those @command{tar} implementations
that are not aware of GNU extensions will at least extract the files
into separate directories, giving the user a possibility to expand it
afterwards @FIXME-ref{how to extract sparse file using third-party
@command{tar}s}.

The resulting @code{GNU.sparse.map} string can be @emph{very} long.
Although POSIX does not impose any limit on the length of a @code{x}
header variable, this possibly can confuse some tars.

@node PAX 1
@appendixsubsec PAX Format, Version 1.0
@UNREVISED{}

The version @code{1.0} of sparse format was introduced with @GNUTAR{}
1.15.92.  Its main objective was to make the resulting file
extractable with little effort even by non-posix aware @command{tar}
implementations.  Starting from this version, the extended header
preceding a sparse member always contains the following variables that
identify the format being used:

@table @code
@item GNU.sparse.major
Major version

@item GNU.sparse.minor
Minor version
@end table

The @code{name} field in @code{ustar} header contains a special name,
constructed using the following pattern:

@smallexample
%d/GNUSparseFile.%p/%f
@end smallexample

The real name of the sparse file is stored in the variable
@code{GNU.sparse.name}.  The real size of the file is stored in the
variable @code{GNU.sparse.realsize}.

The sparse map itself is stored in the file data block, preceding the actual
file data.  It consists of a series of octal numbers of arbitrary length, delimited 
by newlines. The map is padded with nulls to the nearest block boundary.

The first number gives the number of entries in the map. Following are map entries,
each one consisting of two numbers giving the offset and size of the
data block it describes.

The format is designed in such a way that non-posix aware tars and tars not
supporting @code{GNU.sparse.*} keywords will extract each sparse file
in its condensed form with the file map prepended and will place it
into a separate directory.  Then, using a simple program it would be
possible to expand the file to its original form even without GNU tar.
@FIXME-xref{how to extract sparse file using third-party
@command{tar}s}. @FIXME{Write the program and give its URL here}.