On Thu, Apr 28, 2011 at 04:13:17PM +0200, dexen deVries wrote: > > funny thing, my current pendrive (which really is nokia n900 phone) has > abundance of powerful hardware -- cpu, ram and 32GB + microSD block storage. > you got me fantasizing. > > 9p served over usb (either directly, as mentioned recently in somebody's post > about registering usb ids), or over tcp/ip over usb would be cool. if only > various OSes interfaced with that easily. > > another silly, but perhaps doable, approach could be to make the phone serve > either ext2/ext3 or fat, translated on-the-fly from whatever's the underlying > fs. as in, the (say) fat would not reside literally on the device, but > relevant parts would be generated on-demand by the phone in its ram and > presented as a blockdevice over usb mass storage proto. > > in a way, a reverse of typical p9 fileserver -- read files, serve filesystem > image. > > > um, how crazy is that? > Well, if you are going to serve in a format that is distinct from what is actually used on the physical media, why not just provide servers for multiple formants - 9p, nfs, smb etc..
I think there are some network enabled external drives that already do that (maybe not the 9p yet).. Of course if you are backing up filesystems, then you still really need a the media and served formats to be capable of preserving all relevent metadata. I encountered this most recently on a project where I was developing an embedded Linux application for a client whose development environment was exclusively Windows cross development based. I developed on my Linux laptop, but they wanted the kernel source tree checked into their surround SCM. I tried using samba to copy the tree to windows (which hosted the surround client) but of course it failed after about half an hour due to some incompatible file names. Trying to check in a complete vmware image exceeded the maximum file size, and in any case would have defeated the revision management capabilities. Any renaming of files would have broken who knows how many parts of the convoluted set of makefiles and scripts. If you only care about the contents of a modest number of files, interoperability isnt such a big problem. If you want to be able to make accurate backups of entire filesystems then it gets hard if the target is not a natice filesystem format. Regards, DigbyT -- Digby R. S. Tarvin digbyt(at)digbyt.com http://www.digbyt.com
