You are mixing two different problems. An .ist file doesn't influence the final appearance of the index, apart from enabling grouping, for instance. It doesn't have anything to do with the number of columns, which is LaTeX's duty to determine via the definition of theindex (that imakeidx modifies to have, for instance, a different number of columns than two).
Say you want a main index, typeset in two columns, using the main.ist style, and a name index, typeset in three columns, using the names.ist style. Then the following declarations will do:
\usepackage{imakeidx}
\makeindex[title=Main Index,columns=2,options=-s main]
\makeindex[name=names,title=Index of Names,colums=3,options=-s names]
An entry for the main index will be defined by \index{tetrapiloctomy}, while an entry for the name index will be defined by \index[names]{Doe, John}.
In general, the options key tells makeindex or xindy what command line options to use (the input and output file names are determined by the package).
In case you have many indices, it's wise to use the splitindex option to imakeidx (\usepackage[splitindex]{imakeidx}), but this requires running LaTeX with the -shell-escape option or to run separately makeindex.
On 13 March 1996, Don Knuth met with members of the Dutch TeX group (NTG) for a Q&A session.
During this session several questions came up that alluded to troff. The transcript of the entire Q&A session appears in TUGboat 17:4, pp.342-355, with the immediately relevant material starting on p.348. But the underlying philosophy is laid out in the answer to the question from Frans Goddijn that starts on p.347.
After some discussion that mentioned troff, Andries Lenstra asked "Why didn't you start from troff? It was completely inappropriate?" This perhaps doesn't completely answer your question, but the short answer is that troff at that point was layer on layer of patches, and Knuth felt that another patch wasn't the answer -- time to start all over. (After all, as he pointed out a little later, he scrapped TeX78 when he decided that he had found a better approach, and wanted a clean implementation.) It's not so hard to understand when you realize that he's a perfectionist.
Quite a few of Knuth's Q&A sessions were transcribed in TUGboat; to see all of them, look for the heading Knuth interviews and Q&A sessions in the list of contents by keyword.
Best Answer
If you read the Ph.D. thesis of Håkon Wium Lie's you will see that a lot of the concepts of CSS were based on TeX and LateX. Wium Lie and all the other contributor's to CSS came up with a system of separating content from presentation for HTML. The numerous forums and tutorials on the web is an indirect proof that the CSS model is not the best either. You probably find it easier as you are used to it. In addition CSS does not provide any means of programming. This has to be done via JavaScript. It also cannot help with mathematics etc.
When Knuth invented TeX, there was no Computer Science, no web, no html and no CSS. My personal opinion is exactly the opposite. If html was marked as LaTeX life would have been easier, the web would have looked better etc. If there is one aspect that I agree with you is the lack of a
<div></div>concept and the ability to float sections left or right. This can be done using TeX but not as easily. There is an XML version of LaTeX calledtralicsperhaps you can have a look as to how it could have looked!Edit
The current shortcomings of CSS/html can easily be observed by the fact that there are over 500 billion pdf documents on the web and the best way to create a pdf currently is with one of the TeX->pdf engines.