Hello Tim, sorry for my late response Tim Arnold writes: > Hi > I want to create an online doc index using Xindy, > and of course, I need to define a location class > that can accept a URL. > > I understand how to do that for a standard format > of a URL, but I need to be able to handle urls of > arbitrary length. > > My question for the group is: > Where are the *.xdy file(s) that were used to create > the index of the Xindy manual (e.g. manual-6.html, in the > most recent distibution)? This index is the result of hand-weaving the index and it has not been processed by xindy. Sorry for that. Actually the problem you mention is not trivial to solve per se. I don't know how deep you have digged into xindy but one of its assumptions about the ordering of location references is that you can represent a location reference for ordering purposes as a pair of (type, vector of integer). The order of types is specified with the DEFINE-LOCATION-CLASS-ORDER primitive. The order of integer values is used to sort the references lexicographically, e.g. (1 2) < (2) < (3 1) < (3 2), etc. Unfortunately this is hard-wired into the internals of xindy and one should stick to that to avoid rewriting parts of it. Each of the numbers in this vector is derived from the so-called basetypes, e.g. basetype "Roman" gives you '1000' for a 'M'. The problem is that a basetype is only allowed to result in a single number whereas in the URL example one actually needs a whole vector of numbers as response. Ex. Sum.: Currently this does not work! I have some lines of code that could be used to circumvent this problem. Unfortunately this requires to re-dump xindy and I currently have not all the missing pieces together. If you are able to do that I can send you some information on how to apply these patches to do some tests. Cheers, Roger -- ====================================================================== Roger Kehr kehr@informatik.tu-darmstadt.de Computer Science Department Darmstadt University of Technology