On Sun, Jul 25, 1999 at 08:48:25PM -0500, David R. Linn wrote: > >> > | come up with a problem: how to handle file forks in the page cache? > > If you want to resurrect the recent ALBOD discussion (though I prefer > SF, for structured file), you could implement each NTFS "file" as a > Linux directory with each attribute/stream that doesn't correspond > to an inode component as a "file" in the directory (and addtionally > present the entire-SF-as-a-byte-stream as an extra attribute with a > special reserved name. No, I'm pretty strongly against file forks in kernel space, but the problem with NTFS is that they are needed by the filesystem: all the meta-deta is stored in forks. And the problem with presenting them as directories is the same: how to assign inode numbers, when it's quite possible that there won't be enough free (NTFS inode numbers are 48 bit, Linux inumbers are 32 bits on a 32-bit arch, AIUI). -- "It is easier for a camel to pass through the eye of a needle if it is lightly greased." -- Kehlog Albran, "The Profit"
