"Axel D=F6rfler" <axeld@xxxxxxxxxxxxxxxx> wrote: > > "Ingo Weinhold" <bonefish@xxxxxxxxxxxxxxx> wrote: > > > "Ingo Weinhold" <bonefish@xxxxxxxxxxxxxxx> wrote: > > > > > Right, that would be another good idea as long as we don't > > > > > allow > > > > > our > > > > > users to resize partitions and move them around :-) > > > > > (which will be perfectly true for R1) > > > > Even then I see no problem, since the system would update the > > > > data. > > > Of course, that would only be possible when the system knows > > > about > > > all > > > the mount points to be updated. > > > If we don't have that global mount settings file, and the file > > > system > > > can't query reparse points, only the settings of mounted volumes > > > could > > > be updated. > > Mmh, I can't follow here. Since the actual partitioning is done in > > kernel, it should know, which partitions are concerned, even, if > > they > > aren't mounted. Or do I miss something=3D3D3F > > I think so. What I was referring to was this: you have a reparse > point > that creates a point in /boot/home/mail. That reparse point is stored > in the attributes of the real BFS directory /boot/home/mail, but it's > not yet mounted, since the system had no reason to do so yet. > Now, you change the partition that this mount point referred to - if > the reparse point has used the partition size/offset to identify the > partition, the situation that I tried to explain above arises: the > system doesn't necessarily know about the reparse point, and thus, it > can't update it. > Clearer now=3D3F Yes, thanks. I wasn't familar with reparse point idea and where the data for them is stored. If a global settings file on the boot partition would be used, this problem wouldn't occur. This may not work very well with removable media. But on the other hand, if I use a media in another machine, then I this may have undesired effects (e.g. a volume is mounted that should not be mounted). > BTW I think that only the real partition device should be used for > those automatic reparse points - after all, that's the most secure > value since the order and number of partitions rarely changes. > Of course, we could have a global table that maps unique IDs to > partitions, and those can be updated automatically if anything > changes. > But this would also be problematic, if you wanted to use a permanent > / > floppy/ kind of mount point (or in other words, for removable > devices). > > That's why I would say: screw that additional data for system > maintained mount points, and only have /dev/disk/.../master/0=3D5F1 and > the > like as references. Yes. CU, Ingo