How to Scan.

In my experience, it is best to scan directly into a pdf file with the scanner set for black and white. For typed material, 300 dpi is adequate but 400dpi is better; for printed matter, 400dpi is adequate but 600dpi is better.

First method

The best method is to dismantle the book, and feed it into a fast scanner equipped with a sheetfeeder. These are available for about 450USD and will scan about 15 sheets (30 sides) a minute. Dismantling Vol II of AA1988 and scanning it took only about 40 minutes. I used a Fujitsu Scansnap, which works brilliantly except that the feeder is not as robust as I would like --- there are similar models available from Kodak and Xerox. Thus

• Dismantle the book (cut the binding into segments, and trim the binding off the pages in each segment).
• Run it through the scanner.
• Check that all pages are correctly scanned.
• Use acrobat to adjust the pdf page number to agree with that on the pages (display the thumbnails and choose options > number pages).
• Use acrobat to add bookmarks for at least the chapter headings (Ctrl-B etc..)
• Exploit the invariance of the laws of physics under t -> -t to reverse the first step (not tested).

Second method

If you don't want to dismantle the book, you can use a flatbed scanner that allows you to

• scan material up to 8.5in x 11.7in = max{A4,letter},
• at 400dpi using a black and white (text) setting,
• directly into a pdf file.

Most cheap (c 100USD) scanners can do the first two, but not all allow the third. I used a Canon CanoScan 8400F, which works well (except that I wish it were faster, but this is problem with most flatbed scanners, even some that are much more expensive; I think the 8400F is relatively fast). Scanning Vol I of AA1988 took about 80 minutes by this method. If you scan two pages onto one sheet, you need to separate and collate them. This you can do using TeX! Either use Acrobat's crop tool to produce files even.pdf and odd.pdf containing the even and odd pages respectively, and use evenandodd.tex to collate them, or directly use two2one.tex to both split and collate them. The first method is better because it gives you more control.

Third method

Photocopy the material, and then feed the copy into fast sheetfeed scanner. If the copying is done well, the quality can equal that of the second method, and if the copier is fast, it may be as fast as the second method, but I find the second method to be more reliable.

Fourth method

In principle, you should be able to use a digital camera to photograph the pages, but I haven't tried this.

Comments on Copyright and Fair Use Laws

The original US copyright laws protected material for only fifteen years, except that the author was allowed to request one fifteen-year extension. Unfortunately, at the request of a few giant media corporations (for example, the owners of the rights to Donald Duck and Mickey Mouse), the US Congress has repeatedly extended the period until now most works printed after 1922 are protected. This appears to violate the US Constitution, which grants Congress the power only "to promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries." Unfortunately, countries signing "free" trade agreements with the US are often required to adopt the same laws.

Fortunately, there is a tradition that, at the author's request, the publisher will return the copyright to him once the book has gone out of print. Thus it is the responsibility of the author to make sure that his work remains available by putting it on the web. Since this is easy, the only out-of-print works not available on the web should be those whose authors choose not to put it there.

The situation with collections, for example, conference proceedings, is less clear. After some effort, I was able to get permission from Elsevier to post scans of the proceedings of the 1988 Ann Arbor conference on my website, but others have been unsuccessful with Springer. I would urge all editors to make sure that their contract with the publisher contains a provision allowing them to post the work on the web when it goes out of print or after a certain period (whichever occurs first). Clearly, if your purpose in publishing a proceedings is to make the work available to the mathematical community, then it makes no sense to give it to a publisher who will print a few hundred copies at an inflated price, and then block its further distribution for a hundred years.

Once Google completes scanning everything, putting a book on the web will require nothing more than the copyright holder giving Google permission to display the full text. Thus, the only works unavailable on the web will be those whose copyright holder chooses to withhold permission.

Fair use

Under the copyright law of the US (Title 17, US Code), you are generally allowed to copy

• a chapter from a library book,
• a complete work when the work is not available at a "fair price" (this is generally the case when the work is out of print and used copies are not available at a reasonable price),

provided the copy is used only for private study, scholarship, or research.

As far as I know, copying or scanning for private use a book that you legally own, for example, to have it available when travelling, is also considered "fair use" under US law (but see protectfairuse)

For an excellent website on Copyright and Fair Use Law, see Stanford University Library