Reducing everything to a file and using file filters
Maybe part of the solution to the problem that bookmarks cannot easily be integrated into the file system would be to write an http file system (maybe one already exists?). I am thinking you might mount it under say /http and reach web documents by standard URL:s like /http/user.passwd@host:port/path/filename.xyz where ofcourse user, passwd, port and path would be optional. This way you could create symlinks anywhere you want to save your web documents (enabling you to choose any filename you like) that points to somewhere under /http and then use it as a (probably but not necessarily exclusively (I guess?) read-only) file. This way you would always be using the latest version of the file and all tools that works with files would also work transparently with web documents.
Ofcourse you can imagine the http file system automatically finding stale links, doing regular download/backup of documents in the filesystem, etc and ofcourse again there is no reason why other protocols than http couldn't be used the same way.
...and ofcourse another useful protocol to implement a file system for would be standard mail folders, but as I have read the comments to the article there seams to be some mail storage standards for reaching the mail folder as a part of the file system.
Well, reducing bookmarks and mail folders to files might not be what we want, but it would be rather easy to do the indexing etc. for people wanting to be able to search their HOME.
When it comes to searching/indexing .ps and .pdf files and the like I do not understand why the indexer/searcher cannot apply filters based on filename extensions, magic numbers or mime type for example.
Wanting to use text tools like grep on .ps and .pdf files and the like can be solved in at least two ways:
1) make the tools (e.g. grep) smarter and allow for a filter to be applied to the files grep searches before grep does its search
2) make your favourite shell recognize a special charachter that you for example add just before filenames you want to read as plain text at a command line, then the shell applies the filter mechanism (e.g. pdf2txt or ps2txt) and sends the filtered file (probably stored in something like /tmp/shellfiltermechanism/path_to_the_real_file) into the command you entered (for example grep). of course grep would find matches in /tmp/shellfiltermechanism/foo/bar/fie.pdf but you would know you should look at /foo/bar/fie.pdf. This is ofcourse not a perfect solution, but people using grep and the special shell char would know how to handle it, so for us hackers it might do? :)
Just some thoughts...