On 20100425_161747, John Hasler wrote: > Ron Johnson writes: > > You all should (with some research and planning) be able to reorganize > > that sprawling flat structure into a significantly nested structure. > > I wrote: > > Unless it is hard-coded into a closed-source program. > > Ron Johnson writes: > > That scenario can (probably) be worked around with symlinks. > > True. He might also be able to use a unionfs. > -- > John Hasler > > > -- > To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org > with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org > Archive: http://lists.debian.org/87wrvv6xl0....@thumper.dhh.gt.org >
I worked on a personal project last year in which I ran into the 32k limit and solved the problem by imposing a simple structure on my directory names. I wanted to study the frequency of occurance of files with exactly the same content. To do this I set out to compute the md5sum of every ordinary file on four hosts (all running Debian GNU Linux ext3). I wanted to have a directory named with the md5sum of a file. One directory for each ordinary file. In this directory I intended to store information about occurances of that particular file contents. All these files were to be under a root directory that I name filedata/. The first implementation hit a wall at 32998 sub-directories, as most people following this thread now know, but was a total surprise to me at the time. I solved the problem by introducing two levels of sub-directories under filedata/. Instead of having a sub-directory named, for example ed7e26ec800369ca9f694b54e4ed50c4, I introduced into this string two slashes and created the needed directory structure with mkdir -p filedata/ed7/e26/ec800369ca9f694b54e4ed50c4 This resulted, rather quickly as data was collected, in a structure in which there were 4096 sub-directories directly under filedata/ and after more data collection 9 +/-3 sub-directories under each of these first tier sub-directories. In following this thread, I was struck by the difficulty that was caused by poorly devised technical jargon. "Folder" has no history of use in the UNIX/POSIX environment, and with good reason. A directory does not -contain- files in the sense that a manilla folder contains a paper document. It contains information about a name, as in a telephone directory. Furthermore, the original design documents of UNIX place great importance on the fact that a directory -was- -itself- a file, all be it one that was marked for special handling by the OS. Making a tree structure out of files, only, without some alternative structure to handle the tree was a point of some pride for them. They also took some pride in the reasonableness of the names they assigned to the objects they invented. For users, it is unfortunate that the word folder is used for this function, as very few users are unfamiliar with the idea of a telephone directory. At least few are so clueless that they think their girl-friend or boy-friend is actually pressed within the pages of that tome. Folder worked for a while in the early days of the Apple computer where the first file system on floppy disks had only top-level directories (called folders) and no nested directories. I think because Apple didn't want to trouble their users with difficult concepts, like telephones and dialing a telephone number. Who remembers dialing? Where did that word come from? ;-) -- Paul E Condon pecon...@mesanetworks.net -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20100426034739.gb3...@big.lan.gnu
