Interesting to see what happens over the network on that - to windows servers it looks like the 1st write at a high offset (at least for a non-sparse file) causes huge delays - potentially causing the write to timeout. From poking around with the NT IOCTs it looks like they have not only the sparse bit that is a hint that most ranges are unallocated (assumed zero) but useful calls to query the non-zero ranges. There is probably a way to accomplish the same thing as you note.
Steve French Senior Software Engineer Linux Technology Center - IBM Austin phone: 512-838-2294 email: [EMAIL PROTECTED]
