Delivery Formats

Recommendation:

The delivery formats should allow:

• mixed image/text data (multi-layered page structures), not only showing the scanned image, but also carrying (hidden) text layers for OCR text, links and other annotations.

  Explanation: (hidden) annotations like OCR-ed versions of the text should be possible and included to make the documents searchable, Internet links should be possible to let references point to reviews and/or to their location in the DML, and resolve other references, internal and external.

• multi-page faithful images of the scanned raw data

  Explanation: ``Multi-page'' is required to allow convenient access to a whole document, i.e., an article or a book (see the section ``Download Units'' below). ``Faithful'' means: the delivery data should be a lossless compression of the raw scanned data.

• fast downloading, even for large documents (books)

• random access to single pages of documents

  Explanation: It should be possible to look, for example, at the reference page at the end of a 300 page document without downloading all the preceding 299 pages.

• quick rendering, fast sequential page flipping, convenient zooming

  Explanation: This makes the use of the documents convenient and easy. E.g., zooming allows the recognition of small indices and other tiny details.

The delivered files should contain

• Searchable OCR text layers basically in ASCII, non ASCII letters like accents, diaereses in unicode encoding.

• Links to Mathematical Reviews (MR) [[10] ], Zentralblatt für Mathematik (ZBL) [[20] ], Jahrbuch über die Fortschritte der Mathematik, and other similar sources (e.g., there are (or have been) other such services - the Russian Refrativni Zhur. Mat. & the contemporary of the JFM (merged with it in 1933) - Revue Semestrielle des Publications Mathemétiques).

• no ``garbage text'' (e.g., from unrecognized material like formulae)

Further Explanations, Examples:

When choosing a file format, it is important to check that knowledge and implementation of this format is sufficiently stable among digitization and software vendors, or that a sufficient free software community supports it. Proprietary formats, only supported by fragile start-ups and whose conversion or management depend on very specific or lossy software, is to be avoided.

Two file formats currently conform to the requirements above: DjVu [6] , PDF [15].