| Network Working Group | J. Kunze |
| Internet-Draft | California Digital Library |
| Intended status: Informational | J. Littman |
| Expires: November 23, 2018 | George Washington University Libraries |
| E. Summers | |
| University of Maryland | |
| L. Madden | |
| J. Scancella | |
| R. Storey | |
| D. Brunton | |
| K. Zwaard | |
| C. Adams | |
| Library of Congress | |
| A. Boyko | |
| B. Vargas | |
| - | |
| May 22, 2018 |
The BagIt File Packaging Format (V1.0)
draft-kunze-bagit-15
This document specifies BagIt, a set of hierarchical file layout conventions for storage and transfer of arbitrary digital content. A "bag" has just enough structure to enclose descriptive metadata "tags" and a file "payload" but does not require knowledge of the payload's internal semantics. This BagIt format should be suitable for reliable storage and transfer.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on November 23, 2018.
Copyright (c) 2018 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
BagIt is a set of hierarchical file layout conventions designed to support storage and transfer of arbitrary digital content. A bag consists of a directory containing the payload files and other accompanying metadata files known as "tag" files. The "tags" are metadata files intended to facilitate and document the storage and transfer of the bag. Processing a bag does not require any understanding of the payload file contents and the payload files can be accessed without processing the BagIt metadata.
The name, BagIt, is inspired by the "enclose and deposit" method [ENCDEP], sometimes referred to as "bag it and tag it". BagIt differs from serialized archive formats such as MIME, TAR, or ZIP in two general areas:
BagIt is widely used for preserving digital assets originating from different domains. Organizations involved in digital preservation with BagIt include the Library of Congress, Dryad Data Repository, NSF DataONE, and the Rockefeller Archive Center. Software implementations are available for many languages including Python, Ruby, Java, Perl, and PHP. It is also used in the libraries of many universities, such as Cornell, Purdue, Stanford, Ghent University, New York University, and the University of California.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
Implementers are strongly encouraged to review the interoperability considerations described in Section 6.1.
The following terms have precise definitions as used in this specification:
A bag MUST consist of a base directory containing:
The tag files in the base directory consist of one or more files named "manifest-algorithm.txt" (see Section 2.1.3 and Section 2.4), a file named "bagit.txt" (see Section 2.1.1), and zero or more additional tag files (see Section 2.2). The tag files and directories are in arbitrary file hierarchies and MAY have any name that is not reserved for a file or directory in this specification.
The base directory can have any name.
<base directory>/
|
+-- bagit.txt
|
+-- manifest-<algorithm>.txt
|
+-- [optional additional tag files]
|
+-- data/
| |
| +-- [payload files]
|
+-- [optional tag directories]/
|
+-- [optional tag files]
The "bagit.txt" tag file MUST consist of exactly two lines in this order:
BagIt-Version: M.N Tag-File-Character-Encoding: ENCODING
M.N identifies the BagIt major (M) and minor (N) version numbers. ENCODING identifies the character set encoding used by the remaining tag files. ENCODING SHOULD be UTF-8 but for backwards compatibility it MAY be any other encoding registered in [RFC2978]. The bag declaration itself MUST be encoded in UTF-8, and MUST NOT contain a byte-order mark (BOM) [RFC3629].
The number for this version of the specification is "1.0".
The base directory MUST contain a sub-directory named "data".
The payload directory contains the arbitrary digital content within the bag. The files under the payload directory are called payload files, or the payload. Each payload file is treated as an opaque octet stream when verifying file correctness. Payload files MAY be organized in arbitrary sub-directory structures within the payload directory, however for the purpose of this specification such sub-directory structures and filenames have no given meaning.
A payload manifest file provides a complete listing of each payload file along with a corresponding checksum to permit data integrity checking. Manifest entries MUST satisfy the following constraints:
Example payload manifest filenames
manifest-sha256.txt manifest-sha512.txt
Each line of a payload manifest file MUST be of the form:
checksum filepath
where filepath is the pathname of a file relative to the base directory, and checksum is a hex-encoded checksum calculated according to algorithm over every octet in the file.
A manifest MUST NOT reference directories. Bag creators who wish to create an otherwise empty directory have typically done so by creating an empty placeholder file with a name such as ".keep".
A tag manifest is a tag file that lists other tag files and checksums for those tag files generated using a particular bag checksum algorithm.
A bag MAY contain one or more tag manifests, in which case each tag manifest SHOULD list the same set of tag files.
Each tag manifest MUST list every payload manifest. Each tag manifest MUST NOT list any tag manifests, but SHOULD list the remaining tag files present in the bag.
A tag manifest file MUST have a name of the form "tagmanifest-algorithm.txt", where algorithm is a string following the format described in Section 2.4 specifying the bag checksum algorithm used in that manifest.
Tag manifests SHOULD use the same algorithms as the payload manifests that are present in the bag.
Example tag manifest filenames:
tagmanifest-sha256.txt tagmanifest-sha512.txt
A tag manifest file has the same form as the payload file manifest file described in Section 2.1.3, but MUST NOT list any payload files. As a result, no filepath listed in a tag manifest begins "data/".
The "bag-info.txt" file is a tag file that contains metadata elements describing the bag and the payload. The metadata elements contained in the "bag-info.txt" file are intended primarily for human use. All metadata elements are OPTIONAL and MAY be repeated. Because "bag-info.txt" is intended for human reading and editing, ordering MAY be significant and the ordering of metadata elements MUST be preserved.
A metadata element MUST consist of a label, a colon ":", a single linear whitespace character (space or tab), and a value, terminated with a line feed (CR), carriage return (LF) or carriage return plus line feed (CRLF).
The label MUST NOT contain colon (:), line feeds (LF) or carriage returns (CR). The label MAY contain linear whitespace characters, but MUST NOT start or end with whitespace.
It is RECOMMENDED that lines not exceed 79 characters in length. Long values MAY be continued onto the next line by inserting a line feed (LF), a carriage return (CR), or carriage return plus line feed (CRLF) and indenting the next line with one or more linear white space (spaces or tabs). Except for linebreaks such padding does not form part of the value.
Implementations wishing to support previous BagIt versions MUST accept multiple linear whitespace before and after the colon when the bag version is earlier than 1.0; such whitespace does not form part of the label or value.
The following are reserved metadata elements. The use of these reserved metadata elements are OPTIONAL but encouraged. Reserved metadata element names are case-insensitive. Except where indicated otherwise, these metadata element names MAY be repeated to capture multiple values.
In addition to these metadata elements, other arbitrary metadata elements MAY also be present.
An example "bag-info.txt" file
Source-Organization: FOO University
Organization-Address: 1 Main St., Cupertino, California, 11111
Contact-Name: Jane Doe
Contact-Phone: +1 111-111-1111
Contact-Email: example@example.com
External-Description: Uncompressed greyscale TIFF images from the
FOO papers colle...
Bagging-Date: 2008-01-15
External-Identifier: university_foo_001
Payload-Oxum: 279164409832.1198
Bag-Group-Identifier: university_foo
Bag-Count: 1 of 15
Internal-Sender-Identifier: /storage/images/foo
Internal-Sender-Description: Uncompressed greyscale TIFFs created
from microfilm and are...
For reasons of efficiency, a bag MAY be sent with a list of files to be fetched and added to the payload before it can meaningfully be checked for completeness. An OPTIONAL tag file called the fetch file contains such a list.
The fetch file MUST be named "fetch.txt". Every file listed in the fetch file MUST be listed in every payload manifest. A fetch file MUST NOT list any tag files.
Each line of a fetch file MUST be of the form:
url length filepath
where url identifies the file to be fetched and MUST be an absolute URI as defined in [RFC3986], length is the number of octets in the file (or "-", to leave it unspecified), and filepath identifies the corresponding payload file, relative to the base directory.
The slash character ('/') MUST be used as a path separator in filepath. One or more linear whitespace characters (spaces or tabs) MUST separate these three values, and any such characters in the url MUST be percent-encoded [RFC3986]. If filename includes a line feed (LF), a carriage return (CR), carriage return plus line feed (CRLF) or percent sign (%), those characters (and only those) MUST be percent-encoded following [RFC3986]. There is no limitation on the length of any of the fields in the fetch file.
The fetch file allows a bag to be transmitted with "holes" in it, which can be practical for several reasons. For example, it obviates the need for the sender to stage a large serialized copy of the content while the bag is transferred to the receiver. Also, this method allows a sender to construct a bag from components that are either a subset of logically related components (e.g., the localized logical object could be much larger than what is intended for export) or assembled from logically distributed sources (e.g., the object components for export are not stored locally under one filesystem tree).
A bag MAY contain other tag files that are not defined by this specification. Implementations MUST perform standard checksum validation on any tag file which is listed in a tag manifest but MUST otherwise ignore their contents.
All tag files specifically described in this specification MUST adhere to the text tag file format described below. Other tag files MAY adhere to the text tag file format described below.
Text tag files are line-oriented, and each line MUST be terminated by a line feed (LF), a carriage return (CR), or carriage return plus newline (CRLF). It is RECOMMENDED that the last line in a tag file also ends with LF, CR, or CRLF. Text tag file names MUST end in the extension ".txt".
In all text tag files except for the bag declaration file, text MUST be encoded in the character encoding specified in the "bagit.txt" bag declaration file. Text tag files except for the bag declaration file MAY include a byte-order mark (BOM) only if the specified encoding requires it for proper decoding. In accordance with [RFC3629], when "bagit.txt" specifies UTF-8 the tag files MUST NOT begin with a byte-order mark (BOM). See Section 2.1.1
The use of UTF-8 for text tag files is strongly RECOMMENDED. A future version of BagIt may disallow encodings other than UTF-8.
The payload manifest and tag manifests permit validating the integrity of the payload and tag files in a bag produced by the checksum algorithms. Checksum values MUST be encoded so as to conform to the manifest format specified in Section 2.1.3. However, the internal details of a checksum are outside the scope of this document.
To avoid future ambiguity, the checksum algorithm SHOULD be registered in IANA's "Named Information Hash Algorithm Registry" [ni-registry] according to [RFC6920], but MAY for backwards compatibility also be MD5 [RFC1321] or SHA-1 [RFC3174].
The name of the checksum algorithm MUST be normalized for use in the manifest's filename by lowercasing the common name of the algorithm and removing all non-alphanumeric characters. Following is a partial list mapping common algorithm names to normalized names:
Starting with BagIt 1.0, bag creation and validation tools MUST support the SHA-256 and SHA-512 algorithms [RFC6234] and SHOULD enable SHA-512 by default when creating new bags. For backwards compatibility implementers SHOULD support MD5 [RFC1321] and SHA-1 [RFC3174]. Implementers are encouraged to simplify the process of adding additional manifests using new algorithms to streamline the process of in-place upgrades.
A complete bag MUST meet the following requirements:
A valid bag MUST meet the following requirements:
This is the layout of a basic bag containing an image and a companion OCR file. Lines of file content are shown with added parentheses to indicate each complete line. For brevity this example uses MD5 rather than the recommended SHA-512.
myfirstbag/
|
| manifest-md5.txt
| (49afbd86a1ca9f34b677a3f09655eae9 data/27613-h/images/q172.png)
| (408ad21d50cef31da4df6d9ed81b01a7 data/27613-h/images/q172.txt)
|
| bagit.txt
| (BagIt-version: 1.0 )
| (Tag-File-Character-Encoding: UTF-8 )
|
\--- data/
|
| 27613-h/images/q172.png
| (... image bytes ... )
|
| 27613-h/images/q172.txt
| (... OCR text ... )
....
This is the layout of a bag which expects the receiver to download the files listed in the payload manifests prior to validation. Lines of file content are shown with added parentheses to indicate each complete line. For brevity this example uses MD5 rather than the recommended SHA-512.
highsmith-tahoe/ | | manifest-md5.txt | (102b0e6effe208ef9b29864946de9e22 data/23364a.tif ) | | fetch.txt | (https://cdn.loc.gov/master/pnp/highsm/23300/23364a.tif | 216951362 data/23364a.tif ) | | bagit.txt | (BagIt-version: 1.0 ) | (Tag-File-Character-Encoding: UTF-8 ) | | bag-info.txt | (Internal-Sender-Description: Download link found at ) | ( https://www.loc.gov/resource/highsm.23364/ )
The paths specified in the payload manifests, tag manifests, and fetch files do not prohibit special directory characters which have special meaning on some operating systems. Implementers MUST ensure that files outside the bag directory structure are not accessed when reading or writing files based on paths specified in a bag.
All implementations SHOULD have a test suite to guard against special directory characters.
For example, a maliciously crafted "tagmanifest-sha512.txt" file might contain entries which begin with a path character such as "/", "..", or a "~username" home directory reference in an attempt to cause a naive implementation to leak or overwrite targeted files on a POSIX operating system.
Windows implementations SHOULD test their implementations to ensure that safety-checks prevent use of drive letters and the less commonly used namespace sequences (e.g. "\\?\C:\…") described in [MSFNAM].
To assist implementers, the Library of Congress conformance suite [[LC-CONFORMANCE-SUITE]] has some tests for invalid bags which are expected to fail on POSIX or Windows clients.
Implementers of tools that complete bags by retrieving URLs listed in a fetch file need to be aware that some of those URLs might point to hosts, intentionally or unintentionally, that are not under control of the bag's sender. Checksums are intended as a reasonable guarantee against corruption during transit, not a strong cryptographic protection against intentional spoofing.
The size of files, as optionally reported in the fetch file, cannot be guaranteed to match the actual file size to be downloaded. Implementers SHOULD take care to appropriately handle cases where the actual file size does not match the file size reported in the fetch file. Implementers SHOULD NOT use the file size in the fetch file for critical resource allocation, such as buffer sizing or storage requisitioning.
This section lists practical considerations for implementers and users. None of the points below are required but they are recommended for general-purpose usage.
The Library of Congress conformance suite [[LC-CONFORMANCE-SUITE]] is provided as a public resource to test new implementations for compatibility and error handling.
This section provides background information on various challenges caused by differences in how operating systems, filesystems, and common tools handle filenames followed by a list of recommendations for implementers in Section 6.1.1.3.
There are two challenges for interoperability related to filename case:
The Unicode specification has common cases where different character sequences produce the same human-meaningful text. These are referred to as “canonically equivalent” and the Unicode specification defines different normalization forms — see [UNICODE-TR15] for the full details and a brief example below:
The common surname "Núñez" normalized in different forms
Normalization Form D (Decomposition): Char UTF8 Hex Name ---------------------------------------------- N 4e LATIN CAPITAL LETTER N u 75 LATIN SMALL LETTER U \u0301 cc81 COMBINING ACUTE ACCENT n 6e LATIN SMALL LETTER N \u0303 cc83 COMBINING TILDE e 65 LATIN SMALL LETTER E z 7a LATIN SMALL LETTER Z Normalization Form C (Canonical Composition): Char UTF8 Hex Name ---------------------------------------------- N 4e LATIN CAPITAL LETTER N ú c3ba LATIN SMALL LETTER U WITH ACUTE ñ c3b1 LATIN SMALL LETTER N WITH TILDE e 65 LATIN SMALL LETTER E z 7a LATIN SMALL LETTER Z
Unicode normalization is relevant to BagIt implementors because different systems have different standards for normalization:
In practice, this means that the encoded filename stored in a manifest may fail a simple file existence check because the filename's normalization was changed at some point after the manifest was written. This situation is very confusing for users because the filenames are visually indistinguishable and the “missing” file is obviously present in the payload directory.
As specified above, only the Unix-based path separator ('/') may be used inside filenames listed in BagIt manifest and fetch.txt files. When bags are exchanged between Windows and Unix platforms, the path separator SHOULD be translated as needed. Receivers of bags on physical media SHOULD be prepared for filesystems created under either Windows or Unix. Besides the fundamental difference between path separators ('\' and '/'), generally, Windows filesystems have more limitations than Unix filesystems.
Windows path names have a maximum of 255 characters, and none of these characters may be used in a path component:
< > : " / | ? *
Windows also reserves the following names, with or without a file extension:
CON, PRN, AUX, NUL
COM1, COM2, COM3, COM4, COM5, COM6, COM7, COM8, COM9
LPT1, LPT2, LPT3, LPT4, LPT5, LPT6, LPT7, LPT8, LPT9
See [MSFNAM] for more information and possible alternatives.
Some bags have been manually assembled using checksum utilities such as those contained in the GNU Coreutils package (md5sum, sha1sum, etc.), collectively referred to here as "md5sum". Implementers who desire wide support of legacy content should be aware of some known quirks of these tools:
md5sum can be run in “text mode” which causes it to normalize line-endings on some operating systems. On Unix-like systems both modes will usually produce the same results but on systems like Windows they can produce different results based on the file contents. The md5sum output format has two characters between the checksum and the filepath: the first is always a space and the second is an asterisk ("*") for binary mode and a space for text mode.
A final note about md5sum-generated manifests is that for a filepath containing a backslash ('\'), the manifest line will have a backslash inserted in front of the checksum and, under Windows, the backslashes inside filepath can be doubled.
Implementers MAY wish to accept this format by ignoring a leading asterisk or handling differences in line termination gracefully but, if so, implementations MUST warn the user that the bag in question will fail strict validation. In such cases it is RECOMMENDED that tools provide an easy option to update the bag with valid manifests.
The Augmented Backus-Naur form (ABNF) provided below are non-normative. If there is a discrepancy between requirements in the normative sections and the ABNF, the requirements in the normative sections prevail. Some definitions use the core rules (e.g. DIGIT, HEXDIG, etc) as defined in [RFC2234]
bagit.txt ABNF:
bagit-txt = "BagIt-Version: " 1*DIGIT "." 1*DIGIT ending
"Tag-File-Character-Encoding: " encoding ending
encoding = 1*CHAR
ending = CR / LF / CRLF
Payload Manifest ABNF:
payload-manifest = 1*payload-manifest-line
payload-manifest-line = checksum 1*WSP filepath ending
checksum = 1*case-hexdig
case-hexdig = DIGIT / "A" / "a" / "B" / "b" / "C" / "c" /
"D" / "d" / "E"/ "e"/ "F" / "f"
filepath = "data/"
1*( unreserved / pct-encoded / sub-delims )
unreserved = ALPHA / DIGIT / "-" / "." / "_" / "~"
sub-delims = "!" / "$" / "&" / DQUOTE / "'" / "(" / ")" /
"*" / "+" / "," / ";" / "=" / "/"
pct-encoded = "%0D" / "%0d" / "%0A" / "%0a" / "%25"
ending = CR / LF / CRLF
bag-info.txt ABNF:
metadata = 1*metadata-line
metadata-line = key ":" WSP value ending *(continuation ending)
key = 1*non-reserved
value = 1*non-reserved
continuation = WSP 1*non-reserved
non-reserved = VCHAR / WSP
; any valid character for the specific encoding
; except those that match "ending"
ending = CR / LF / CRLF
fetch.txt ABNF:
fetch = 1*fetch-line
fetch-line = url 1*WSP length 1*WSP filepath ending
url = <absolute-URI, see [RFC3986], Section 4.3>
length = 1*DIGIT / "-"
filepath = ("data/"
1*( unreserved / pct-encoded / sub-delims ))
ending = CR / LF / CRLF
BagIt owes much to many thoughtful contributors and reviewers, including Stephen Abrams, Mike Ashenfelder, Dan Chudnov, Dave Crocker, Brad Hards, Scott Fisher, Keith Johnson, Erik Hetzner, Leslie Johnston, David Loy, Mark Phillips, Tracy Seneca, Brian Tingle, Adam Turoff, Jim Tuttle, and Stian Soiland-Reyes.
This draft does not request any action from IANA.