Vector version of Image:Layout_engine_usage.png on English Wikipedia. Original description:

Comparison of layout engine/web browser usage share (market share) over time. Data is based on usage share of web browsers. Note that the values may contain errors. Existing problems in the current data:

1. Data for the early part of the (first) browser wars could use some refinement.
2. Extremely scant data prior to 1996.
3. No statistical data prior to 1994 (possible to infer certain things though...)
4. Very few statistics about Safari and no data about Konqueror

Please help improving the data if you can. :-)

Suggestions for future improvements:

1. Include partial-year data points (see following suggestions :) rather than the rough way of doing it like now.
2. Start of the graph: WorldWideWeb, "the world's first web browser [...] was completed on Christmas Day, 1990 [and was] released to the public (by way of Internet newsgroups) in August 1991."
3. Mosaic "was released on April 22, 1993"; as per usage share of web browsers, by January 1994 it was at 97% (!!!) so a very rapid replacement...
4. "Beta versions of Netscape Navigator were freely downloadable in mid- to late-1994, and version 1.0 of the browser was released by the end of the year." If we had some 1995 data, that would help fill in the blanks, but by October 1994 (as per the usage share of web browsers page again), the usage share for NN was already 18%! (the internet pioneers were certainly early adopters, ne? ;) and by 1996 was well in the majority.
5. "Microsoft licensed Spyglass Enhanced Mosaic as the basis of Internet Explorer 1.0 which it released as part of the Microsoft Windows 95 Plus Pack in August 1995. Internet Explorer 2.0 was released three months later, and by then the race was on." "Internet Explorer was not widely used until version 4, which was released in October 1997."
6. Present data in such a way that the year marker indicates data for January 1 of that year (e.g. reading up from "2005" should indicate approximate market share for Jan. 1, 2005).
7. If you have a better chart drawing tool, please create a better one.

(Original) Graph created with OpenOffice.org Calc. (Recreated in vector form in Inkscape.)

The revised CSV data:

 "Date","Mosaic","Netscape Classic","Trident (Internet Explorer for Windows)","Gecko (Mozilla, Firefox, Netscape 6+, etc)","WebKit (Safari)","Presto (Opera 7+)","Others"  1994,95,0,0,0,0,0,5  1995,37.5,55,0,0,0,0,7.5  1996,7.5,80,3.75,0,0,0,8.75  1997,2.5,72.5,17.5,0,0,0,7.5  1998,0,60,35,0,0,0,5  1999,0,35,62.5,0,0,0,2.5  2000,0,15,83.75,0,0,0,1.25  2001,0,9,90,0,0,0,1  2002,0,2.5,95.25,1,0,0,1.25  2003,0,0.75,94.75,2,0,0,2.5  2004,0,0,92.625,4,1,0.625,1.75  2005,0,0,86.903861,9.544639,1.571667,0.7429,1.23693299999999  2006,0,0,60.65,9.5925,30.995,0.4825,0.4825 

Some further notes on the data:

1. Because these (especially the very early data) are rough estimates, we should see nice round numbers (or at least numbers that make the graph pretty ;)
2. "WorldWideWeb" currently has only "0" values for data points; I am thus excluding it from the CSV info and graph barring new information.
3. The data for each year should add up to 100 (percent)
4. According to the Mosaic page, "Mosaic's popularity as a separate browser began to dry up upon the release of Netscape Navigator, and by 1998 its userbase had almost completely evaporated." which means it should have non-zero values up until then.
5. "In January 2003, Opera 7 was released and introduced a new layout engine "Presto"." Thus for our graph, the 2003 value (Jan. 1) for Opera (Presto) should be zero; changed accordingly.
6. Since Netscape Browser uses IE/Moz dual engine, unless it is used more widely (which is quite less likely to happen), we should exclude it to avoid the confusion (the contribution from Netscape 8 right now (<1%) would be less than the difference (~3%) between the various sources reporting Firefox's share 8-)