NY Times today (26 June 2015) says Nikola Alic and others have successfully sent optical signals thru an optical fiber 7456 miles long, WITHOUT any "repeaters" along the way. Supposedly, the method involves understanding what kinds of transformations happen to the signal during transit, and pre-applying an approximately-inverse transformation. That causes the signal to be maximally easy to detect at the receiving end. E. Temprana, E. Myslivets, B.P.-P. Kuo, L. Liu, V. Ataie, N. Alic, S. Radic: Overcoming Kerr-induced capacity limit in optical fiber transmission Science 348,6242 (26 June 2015) 1445-1448 http://www.sciencemag.org/content/348/6242/1445 As far as I understand (this may have been hype, but probably not) semiconductor chip manufacturers trying to optically-lithograph ultrafine patterns, have for years all done a similar trick. That is, (a) light is transmitted thru an optical mask defining the pattern, then (b) hits photoresist on the chip, whch (c) polymerizes, then (d) chemical etching or whatever is used to generate the pattern on the chip. This process abcd causes the pattern you get to differ from the pattern on the mask via some transformation which is somewhat understood, and can be, and is, approximately inverted. Hence the picture they draw on the mask intentionally is NOT the same as the intended pattern on the chip. This is actually a harder mathematical problem since 2 or 3 or 4 dimensional, whereas the fiber problem presumably is only 1 or 2 dimensional.