On 3 May 2015, at 9:09pm, Scott Doctor <scott at scottdoctor.com> wrote:
> Hmm, one for doing my own locking, one against it. As this seems to be an > obvious issue in any network, I wonder why the network developers have not > appropriately addressed this issue. They've tried. With modern operating systems locking is part of the network file system -- things like NFS, CIFS/SMB and AFP. And these depend a great deal on three other subsystems: the operating system, the disk file system (FAT, NTFS, HFS) and the network system (probably TCP/IP for all of us). Every combination of these has to be tested, with multiple combinations of different numbers of computers. There are a huge number of combinations and with new drivers and versions arising daily it takes a plan of continuous testing, not just a project with an end-date. I have high hopes for SMB3. It was designed recently, with modern issues and capabilities in mind. Maybe it'll be so trustworthy that Dr Hipp can state that SQLite works fine with it. And Apple's AFP+ is long past its Best By date, though Apple shows signs of having adopted SMB2 for networking already, which suggests it will not bother with its own replacement for AFP+ and just call SMB2 or SMB3 an industry standard. Maybe it's time for an Open, Free (as in beer, with all patents available to all), crowd-developed networking file system, designed with modern issues (e.g. attacks, encryption, transactions, ACID, massive parallelism) in mind. Simon.
