The assets of digitization in "text format" are numerous. It makes a smaller and more easily sendable computer file, unlike digitization in "image format", which produces a bulky "photo" file. Contrary to other formats, the files are accessible for low-bandwidth use. They can be copied as much as needed to produce new digital or print versions for free. The typos pointed out after the text is released can be fixed at any time. Readers can change the font and size of characters, the margins or the number of lines per page. Visually impaired readers can increase the letter size. Blind readers can use speech recognition software. All this is very difficult, if not impossible, with many other formats. If the eBooks released are 99.9% accurate in the eyes of the general reader, the goal is not to create authoritative editions, and to argue with a picky reader whether a certain sentence should have a colon instead of a semi-colon between its clauses. Project Gutenberg is convinced that proofreading by human beings is a very important step, and that this step makes all the difference. The use of scanned books as is --converted to text format by OCR software with no proofreading-- gives a much lower quality result. After running OCR software, the text is 99% reliable, in the best of cases. After proofreading, the text becomes 99.95% reliable (a high percentage which is also the standard at the Library of Congress). For this reason, Project Gutenberg's perspective is rather different from that of the Million Book Project, another project launched by several professors from Carnegie Mellon University, and whose collections (10,611 books on June 1st, 2005) are hosted by the Internet Archive (the Internet Archive is also the backup distribution site of Project Gutenberg). In the case of the Million Book Project, books are scanned and "OCRized", but they are not proofread. The main formats used are XML, TIF and DjVu. On Project Gutenberg's website, a File Recode Service allows users to convert books in one format (ASCII, ISO-8859, Unicode and Big-5) into another, and vice versa. A much more powerful conversion program may be launched in the future, with a conversion into still more formats (XML, HTML, PDF, TeX, RTF), including Braille and voice. It will then also be possible to choose the font and size of characters and the background color. Another eagerly expected conversion is that of a book from one language to another by machine translation software. This may be possible in a few years, when machine translation is accurate to 99%. 5. DISTRIBUTED PROOFREADERS, TO HANDLE SHARED PROOFREADING The main "leap forward" of Project Gutenberg in the last few years is due to Distributed Proofreaders.