Last modified: 2007-10-27 21:20:11 UTC
As far as I can see the current implementation of the LaTeX rendering of mathematical formulas creates image files using the md5 hash of the latex string as filename. Whenever a formula needs to be shown on a website the md5 hash of its latex string is calculated and the filesystem is being checked for a file of that name. If no such file exists it will be created, otherwise the existing image file will be shown. There is an astronomical but non-zero chance of two different formulas having the same md5 hash, resulting in a wrong presentation of the second formula. Even considering the birthday-paradox (http://en.wikipedia.org/wiki/Birthday_paradox) the chances should be rather small but maybe there's someone who nevertheless wants to fix this.
How about using {$pageid}-{$n} as a key, i.e. 1701-0 for the first formula on pageid 1701, 1701-1 for the second, etc. The drawback is that when a formula is added on top of a formula-rich article, the entire article's cache has to be purged (or moved?)
