Network Working Group J. Kunze Request for Comments: 2731 Dublin Core Category: Informational Metadata Initiative December 1999 Encoding Dublin Core Metadata in HTML Status of this Memo This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited. Copyright Notice Copyright (C) The Internet Society (1999). All Rights Reserved. 1. Abstract The Dublin Core [DC1] is a small set of metadata elements for describing information resources. This document explains how these elements are expressed using the META and LINK tags of HTML [HTML4.0]. A sequence of metadata elements embedded in an HTML file is taken to be a description of that file. Examples illustrate conventions allowing interoperation with current software that indexes, displays, and manipulates metadata, such as [SWISH-E], [freeWAIS-sf2.0], [GLIMPSE], [HARVEST], [ISEARCH], etc., and the Perl [PERL] scripts in the appendix. 2. HTML, Dublin Core, and Non-Dublin Core Metadata The Dublin Core (DC) metadata initiative [DCHOME] has produced a small set of resource description categories [DC1], or elements of metadata (literally, data about data). Metadata elements are typically small relative to the resource they describe and may, if the resource format permits, be embedded in it. Two such formats are the Hypertext Markup Language (HTML) and the Extensible Markup Language (XML); HTML is currently in wide use, but once standardized, XML [XML] in conjunction with the Resource Description Framework [RDF] promise a significantly more expressive means of encoding metadata. The [RDF] specification actually describes a way to use RDF within an HTML document by adhering to an abbreviated syntax. Kunze Informational [Page 1] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 This document explains how to encode metadata using HTML 4.0 [HTML4.0]. It is not concerned with element semantics, which are defined elsewhere. For illustrative purposes, some element semantics are alluded to, but in no way should semantics appearing here be considered definitive. The HTML encoding allows elements of DC metadata to be interspersed with non-DC elements (provided such mixing is consistent with rules governing use of those non-DC elements). A DC element is indicated by the prefix "DC", and a non-DC element by another prefix; for example, the prefix "AC" is used with elements from the A-Core [AC]. 3. The META Tag The META tag of HTML is designed to encode a named metadata element. Each element describes a given aspect of a document or other information resource. For example, this tagged metadata element, says that Homer Simpson is the Creator, where the element named Creator is defined in the DC element set. In the more general form, the capitalized words are meant to be replaced in actual descriptions; thus in the example, ELEMENT_NAME was: Creator ELEMENT_VALUE was: Simpson, Homer and PREFIX was: DC Within a META tag the first letter of a Dublin Core element name is capitalized. DC places no restriction on alphabetic case in an element value and any number of META tagged elements may appear together, in any order. More than one DC element with the same name may appear, and each DC element is optional. The next example is a book description with two authors, two titles, and no other metadata. Kunze Informational [Page 2] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 The prefix "DC" precedes each Dublin Core element encoded with META, and it is separated by a period (.) from the element name following it. Each non-DC element should be encoded with a prefix that can be used to trace its origin and definition; the linkage between prefix and element definition is made with the LINK tag, as explained in the next section. Non-DC elements, such as Email from the A-Core [AC], may appear together with DC elements, as in This example also shows how some special characters may be encoded. The author name in the first element contains a diacritic encoded as an HTML character entity reference -- in this case an accented letter E. Similarly, the last line contains two double-quote characters encoded so as to avoid being interpreted as element content delimiters. 4. The LINK Tag The LINK tag of HTML may be used to associate an element name prefix with the reference definition of the element set that it identifies. A sequence of META tags describing a resource is incomplete without one such LINK tag for each different prefix appearing in the sequence. The previous example could be considered complete with the addition of these two LINK tags: In general, the association takes the form Kunze Informational [Page 3] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 where, in actual descriptions, PREFIX is to be replaced by the prefix and LOCATION_OF_DEFINITION by the URL or URN of the defining document. When embedded in the HEAD part of an HTML file, a sequence of LINK and META tags describes the information in the surrounding HTML file itself. Here is a complete HTML file with its own embedded description. A Dirge
               Rough wind, that moanest loud
                 Grief too sad for song;
               Wild wind, when sullen cloud
                 Knells all the night long;
               Sad storm, whose tears are vain,
               Bare woods, whose branches strain,
               Deep caves and dreary main, -
                 Wail, for the world's wrong!
       
5. Encoding Recommendations HTML allows more flexibility in principle and in practice than is recommended here for encoding metadata. Limited flexibility encourages easy development of software for extracting and processing metadata. At this early evolutionary stage of internet metadata, easy prototyping and experimentation hastens the development of useful standards. Kunze Informational [Page 4] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Adherence is therefore recommended to the tagging style exemplified in this document as regards prefix and element name capitalization, double-quoting (") of attribute values, and not starting more than one META tag on a line. There is much room for flexibility, but choosing a style and sticking with it will likely make metadata manipulation and editing easier. The following META tags adhere to the recommendations and carry identical metadata in three different styles: Use of these recommendations is known to result in metadata that may be harvested, indexed, and manipulated by popular, freely available software packages such as [SWISH-E], [freeWAIS-sf2.0], [GLIMPSE], [HARVEST], and [ISEARCH], among others. These conventions also work with the metadata processing scripts appearing in the appendix, as well as with most of the [DCPROJECTS] applications referenced from the [DCHOME] site. Software support for the LINK tag and qualifier conventions (see the next section) is not currently widespread. Ordering of metadata elements is not preserved in general. Writers of software for metadata indexing and display should try to preserve relative ordering among META tagged elements having the same name (e.g., among multiple authors), however, metadata providers and searchers have no guarantee that ordering will be preserved in metadata that passes through unknown systems. 6. Dublin Core in Real Descriptions In actual resource description it is often necessary to qualify Dublin Core elements to add nuances of meaning. While neither the general principles nor the specific semantics of DC qualifiers are within scope of this document, everyday uses of the qualifier syntax are illustrated to lend realism to later examples. Without further explanation, the three ways in which the optional qualifier syntax is currently (subject to change) used to supplement the META tag may be summarized as follows: Kunze Informational [Page 5] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Accordingly, a posthumous work in Spanish might be described with Note that the qualifier syntax and label suffixes (which follow an element name and a period) used in examples in this document merely reflect current trends in the HTML encoding of qualifiers. Use of this syntax and these suffixes is neither a standard nor a recommendation. 7. Encoding Dublin Core Elements This section consists of very simple Dublin Core encoding examples, arranged by element. Title (name given to the resource) ----- Kunze Informational [Page 6] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Creator (entity that created the content) ------- Subject (topic or keyword) ------- Kunze Informational [Page 7] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Description (account, summary, or abstract of the content) ----------- Publisher (entity that made the resource available) --------- Contributor (other entity that made a contribution) ----------- Kunze Informational [Page 8] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Date (of an event in the life of the resource; [WTN8601] recommended) ---- Type (nature, genre, or category; [DCT1] recommended) ---- Kunze Informational [Page 9] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Format (physical or digital data format, plus optional dimensions) ------ Kunze Informational [Page 10] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Identifier (of the resource) ---------- Source (reference to the resource's origin) ------ Language (of the content of the resource; [RFC1766] recommended) -------- Kunze Informational [Page 11] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Relation (reference to a related resource) -------- Coverage (extent or scope of the content) -------- Kunze Informational [Page 12] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Rights (text or identifier of a rights management statement) ------ 8. Security Considerations The syntax rules for encoding Dublin Core metadata in HTML that are documented here pose no direct risk to computers and networks. People can use these rules to encode metadata that is inaccurate or even deliberately misleading (creating mischief in the form of "index spam"), however, this reflects a general pattern of HTML META tag abuse that is not limited to the encoding of metadata from the Dublin Core set. Even traditional metadata encoding schemes (e.g., [MARC]) are not immune to inaccuracy, although they are generally followed in environments where production quality greatly exceeds that of the average Web site. Systems that process metadata encoded with META tags need to consider issues related to its accuracy and validity as part of their design and implementation, and users of such systems need to consider the design and implementation assumptions. Various approaches may be relevant for certain applications, such as adding statements of metadata provenance, signing of metadata with digital signatures, and automating certain aspects of metadata creation; but these are far outside the scope of this document and the underlying META tag syntax that it describes. Kunze Informational [Page 13] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 9. Appendix -- Perl Scripts that Manipulate HTML Encoded Metadata This section contains two simple programs that work with versions 4 and 5 of the Perl [PERL] scripting language interpreter. They may be taken and freely adapted for local organizational needs, research proposals, venture capital bids, etc. A variety of applications are within easy reach of implementors that choose to build on these scripts. Script 1: Metadata Format Conversion ------------------------------------- Here is a simple Perl script that correctly recognizes every example of metadata encoding in this document. It shows how a modest scripting effort can produce a utility that converts metadata from one format to another. Minor changes are sufficient to support a number of output formats. #!/depot/bin/perl # # This simple perl script extracts metadata embedded in an HTML file # and outputs it in an alternate format. Issues warning about missing # element name or value. # # Handles mixed case tags and attribute values, one per line or spanning # several lines. Also handles a quoted string spanning multiple lines. # No error checking. Does not tolerate more than one ") { next if (! //i) { while (<>) { $meta .= $_; last if (/>/); } } $name = $meta =~ /name\s*=\s*"([^"]*)"/i ? $1 : "MISSING ELEMENT NAME"; $content = $meta =~ /content\s*=\s*"([^"]*)"/i ? $1 : "MISSING ELEMENT VALUE"; ($scheme) = $meta =~ /scheme\s*=\s*"([^"]*)"/i; ($lang) = $meta =~ /lang\s*=\s*"([^"]*)"/i; if ($lang || $scheme) { $mod = " ($lang"; if (! $scheme) Kunze Informational [Page 14] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 { $mod .= ")"; } elsif (! $lang) { $mod .= "$scheme)" } else { $mod .= ", $scheme)"; } } else { $mod = ""; } print " @|$name$mod; $content\n"; } print "@)urc;\n"; # ---- end of Perl script ---- When the conversion script is run on the metadata file example from the LINK tag section (section 4), it produces the following output. @(urc; @|DC.Title; A Dirge @|DC.Creator; Shelley, Percy Bysshe @|DC.Type; poem @|DC.Date; 1820 @|DC.Format; text/html @|DC.Language; en @)urc; Script 2: Automated Metadata Creation -------------------------------------- The creation and maintenance of high-quality metadata can be extremely expensive without automation to assist in processes such as supplying pre-set or computed defaults, validating syntax, verifying value ranges, spell checking, etc. Considerable relief could be had from a script that reduced an individual provider's metadata burden to just the title of each document. Below is such a script. It lets the provider of an HTML document abbreviate an entire embedded resource description using a single HTML comment statement that looks like Our script processes this statement specially as a kind of "metadata block" declaration with attached title. The general form is Kunze Informational [Page 15] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 This statement works much like a "Web server-side include" in that the script replaces it with a fully-specified block of metadata and triggers other replacements. Once installed, the script can output HTML files suitable for integration into one's production Web server procedures. The individual provider keeps a separate "template" file of infrequently changing pre-set values for metadata elements. If the provider's needs are simple enough, the only element values besides the title that differ from one document to the next may be generated automatically. Using the script, values may be referenced as variables from within the template or within the document. Our variable references have the form "(--mbVARNAME)", and here is what they look like inside a template: (--mbtitle) The above template represents the metadata block that will describe the document once the variable references are replaced with real values. By the conventions of our script, the following variables will be replaced in both the template and in the document: (--mbfilesize) size of the final output file (--mbtitle) title of the document (--mblanguage) language of the document (--mbbaseURL) beginning part of document identifier (--mbfilename) last part (minus .html) of identifier (--mbfilemodtime) last modification date of the document Kunze Informational [Page 16] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 Here's an example HTML file to run the script on.

From: Acting Shift Supervisor To: Plant Control Personnel RE: (--mbtitle) Date: (--mbfilemodtime)

Pursuant to directive DOH:10.2001/405aec of article B-2022, subsection 48.2.4.4.1c regarding staff morale and employee productivity standards, the current allocation of doughnut acquisition funds shall be increased effective immediately. Note that because replacement occurs throughout the document, the provider need only enter the title once instead of twice (normally the title must be entered once in the HTML head and again in the HTML body). After running the script, the above file is transformed into this: Nutritional Allocation Increase

From: Acting Shift Supervisor To: Plant Control Personnel RE: Nutritional Allocation Increase Date: 1999-03-08

Pursuant to directive DOH:10.2001/405aec of article B-2022, subsection 48.2.4.4.1c regarding staff morale and employee productivity standards, the current allocation of doughnut acquisition funds shall be increased effective immediately. Here is the script that accomplishes this transformation. #!/depot/bin/perl # # This Perl script processes metadata block declarations of the form # and variable references of the # form (--mbVARNAME), replacing them with full metadata blocks and # variable values, respectively. Requires a "template" file. # Outputs an HTML file. # # Invoke this script with a single filename argument, "foo". It creates # an output file "foo.html" using a temporary working file "foo.work". # The size of foo.work is measured after variable replacement, and is # later inserted into the file in such a way that the file's size does # not change in the process. Has little or no error checking. $infile = shift; open(IN, "< $infile") or die("Could not open input file \"$infile\""); $workfile = "$infile.work"; unlink($workfile); open(WORK, "+> $workfile") or die("Could not open work file \"$workfile\""); @offsets = (); # records locations for late size replacement $title = ""; # gets the title during metablock processing $language = "en"; # pre-set language here (not in the template) $baseURL = "http://moes.bar.com/doh"; # pre-set base URL here also $filename = "$infile.html"; # final output filename $filesize = "(--mbfilesize)"; # replaced late (separate pass) Kunze Informational [Page 18] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 ($year, $month, $day) = (localtime( (stat IN) [9] ))[5, 4, 3]; $filemodtime = sprintf "%s-%02s-%02s", 1900 + $year, 1 + $month, $day; sub putout { # outputs current line with variable replacement if (! /\(--mb/) { print WORK; return; } if (/\(--mbfilesize\)/) # remember where it was { push @offsets, tell WORK; } # but don't replace yet s/\(--mbtitle\)/$title/g; s/\(--mblanguage\)/$language/g; s/\(--mbbaseURL\)/$baseURL/g; s/\(--mbfilename\)/$filename/g; s/\(--mbfilemodtime\)/$filemodtime/g; print WORK; } while () { # main loop for input file if (! /(.*)(.*)//) { $remainder = $1; } else { while () { $title .= $_; last if (/(.*)\s*-->(.*)/); } $title .= $1; $remainder = $2; } open(TPLATE, "< template") or die("Could not open template file"); while () # subloop for template file { &putout; } close(TPLATE); $_ = $remainder; &putout; Kunze Informational [Page 19] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 } close(IN); # Now replace filesize variables without altering total byte count. select( (select(WORK), $| = 1) [0] ); # first flush output so we if (($size = -s WORK) < 100000) # can get final file size { $scale = 0; } # and set scale factor or else { # compute it, keeping width of size field low for ($scale = 0; $size >= 1000; $scale++) { $size /= 1024; } } $filesize = sprintf "%7.7s %sbytes", $size, (" ", "K", "M", "G", "T", "P") [$scale]; foreach $pos (@offsets) { # loop through saved size locations seek WORK, $pos, 0; # read the line found there $_ = ; # $filesize must be exactly as wide as "(--mbfilesize)" s/\(--mbfilesize\)/$filesize/g; seek WORK, $pos, 0; # rewrite it with replacement print WORK; } close(WORK); rename($workfile, "$filename") or die("Could not rename \"$workfile\" to \"$filename\""); # ---- end of Perl script ---- Kunze Informational [Page 20] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 10. Author's Address John A. Kunze Center for Knowledge Management University of California, San Francisco 530 Parnassus Ave, Box 0840 San Francisco, CA 94143-0840, USA Fax: +1 415-476-4653 EMail: jak@ckm.ucsf.edu 11. References [AAT] Art and Architecture Thesaurus, Getty Information Institute. http://shiva.pub.getty.edu/aat_browser/ [AC] The A-Core: Metadata about Content Metadata, (in progress) http://metadata.net/ac/draft-iannella-admin-01.txt [DC1] Weibel, S., Kunze, J., Lagoze, C. and M. Wolf, "Dublin Core Metadata for Resource Discovery", RFC 2413, September 1998. ftp://ftp.isi.edu/in-notes/rfc2413.txt [DCHOME] Dublin Core Initiative Home Page. http://purl.org/DC/ [DCPROJECTS] Projects Using Dublin Core Metadata. http://purl.org/DC/projects/index.htm [DCT1] Dublin Core Type List 1, DC Type Working Group, March 1999. http://www.loc.gov/marc/typelist.html [freeWAIS-sf2.0] The enhanced freeWAIS distribution, February 1999. http://ls6-www.cs.uni- dortmund.de/ir/projects/freeWAIS-sf/ [GLIMPSE] Glimpse Home Page. http://glimpse.cs.arizona.edu/ [HARVEST] Harvest Web Indexing. http://www.tardis.ed.ac.uk/harvest/ Kunze Informational [Page 21] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 [HTML4.0] Hypertext Markup Language 4.0 Specification, April 1998. http://www.w3.org/TR/REC-html40/ [ISEARCH] Isearch Resources Page. http://www.etymon.com/Isearch/ [ISO639-2] Code for the representation of names of languages, 1996. http://www.indigo.ie/egt/standards/iso639/iso639-2- en.html [ISO8601] ISO 8601:1988(E), Data elements and interchange formats -- Information interchange -- Representation of dates and times, International Organization for Standardization, June 1988. http://www.iso.ch/markete/8601.pdf [MARC] USMARC Format for Bibliographic Data, US Library of Congress. http://lcweb.loc.gov/marc/marc.html [PERL] L. Wall, T. Christiansen, R. Schwartz, Programming Perl, Second Edition, O'Reilly, 1996. [RDF] Resource Description Framework Model and Syntax Specification, February 1999. http://www.w3.org/TR/REC-rdf-syntax/ [RFC1766] Alvestrand, H., "Tags for the Identification of Languages", RFC 1766, March 1996. ftp://ftp.isi.edu/in-notes/rfc1766.txt [SWISH-E] Simple Web Indexing System for Humans - Enhanced. http://sunsite.Berkeley.EDU/SWISH-E/ [TGN] Thesaurus of Geographic Names, Getty Information Institute. http://shiva.pub.getty.edu/tgn_browser/ [WTN8601] W3C Technical Note - Profile of ISO 8601 Date and Time Formats. http://www.w3.org/TR/NOTE-datetime [XML] Extensible Markup Language (XML). http://www.w3.org/TR/REC-xml Kunze Informational [Page 22] RFC 2731 Encoding Dublin Core Metadata in HTML December 1999 12. Full Copyright Statement Copyright (C) The Internet Society (1999). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Acknowledgement Funding for the RFC Editor function is currently provided by the Internet Society. Kunze Informational [Page 23]