In Dec 1997, J. L. Corey, Director of the Florida Center of Library Automation, provided these estimates of costs, which are based on current costs at FCLA for making about 500,000 MB of information Web-accessible:

• Direct data storage

  About $0.10/MB/year.
  (This assumes a 4 year lifespan of a disk drive and/or technology)

• CPU costs

  About $0.10/MB/year.
  (This assumes 5 year lifespan of a unix cpu/ sp2 node, and t hat 100GB of memory can easily be accessed by one box. Average price of an SP2 thin node with memory, switch card, etc. is $50,000.)

• Personnel

  About $0.15/MB/year.
  (This assumes 2 FTE @ $30,000 + 25% fringe. These 2 FTE, costing $75,000/year) currently tend about 500 GB of files.)

For two reasons these estimates are probably high relative to what future costs will be: (1) As the amount of information made accessible increases, the personnel costs per MB should decrease. (2) The costs of hardware to make information Web-accessible has been steadily decreasing and seems likely to continue to do so.

According to Tenopir and King (1997), the average U.S. scholarly science journal in 1995 had 123 articles and 1723 pages. (Only 1434 of these pages were “article pages”, but to be safe, I used 14 rather than 12 pages for mean article length.)

An average 14-page article produces a PDF file of about 0.6 MB. [I downloaded the PDF files of three 7-page articles in the 5 Dec 1997 issue of Journal of Biological Chemistry (http://www.jbc.org/). Their sizes were 232, 442, and 234 KB (mean=303). Assuming that a 14-page article makes a PDF file twice as large as a 7-page article, I arrived at 0.6 MB.

123 articles x 0.6 MB= 74 MB for 1 year of an average journal

______________

Tenopir, C., and D. W. King. 1997. Trends in scientific scholarly journal publishing in the United States. J. Scholarly Pub. 28 (3): 135-170.