Directory structure (file system design)

I was looking at how file systems are designed and noticed that most places say that the directory hierarchy can be implemented using a hash table.

Is there an established convention for separating Windows file names in a string?

I have a function which needs to output a string containing a list of file paths. I can choose the separation character but I cannot change the data type (e.g. I cannot return a List<string> or something like that).

File system implementation in MongoDB with GridFS

I am working on two projects that will both implement a Webdav server backed by a MongoDB GridFS. In each case, there is the potential for the system to store tens of millions of files spread across thousands of hierarchical directories.

NoSQL as file meta database

I am trying to implement a virtual file system structure in front of an object storage (Openstack). For availability reasons we initially chose Cassandra, however while designing file system data model, it looked like a tree structure similar to a relational model. Here is the dilemma for availability and partition tolerance we need NoSQL, but our data model is relational.

Which language is more suitable heavy file tasks?

I need to write a script (based on basic functions) to process /image/audio/video files. The process is mainly filesystem tasks and converts. The database of files has been stored by mysql. The script is simple but cause heavy tasks on the system; for example renaming/converting/copying thousands of file in a run. The script does not read the content of files into memory, it just manage the commands for sub-processes. The main weight is on the communication with filesystem. The script will be used regularly for new files. My concern is about performance. I am thinking of

Lowercase in Linux file names

As I find that UpperCase are really readable for first letter word separation in long complex names, I tend to give some of my Linux files names with some UpperCase. Mostly executables, some directories too.

Best practice to sync long paths to filesystems where path length is limited?

I wrote a sync tool to synchronize folders/files from Alfresco to Windows.
PROBLEM:

Is it plausible to use a filesystem-based configuration format rather than an INI file?

After having had endless issues with INI-style configuration files and parsing them correctly (let along getting binary data right), a crazy idea crossed my mind recently:

How to detect the encoding of a file?

On my filesystem (Windows 7) I have some text files (These are SQL script files, if that matters).

Is it a bad idea to sync file system with remote server using HTTP?

I have started a project which will duplicate Dropbox or Google Drive behavior but using Amazon S3 az a backend.