[Techtalk] finding bad files

[Wim De Smet]

On Wed, Apr 15, 2015 at 7:33 PM John Sturdy <jcg.sturdy at gmail.com> wrote:

> On Mon, Apr 13, 2015 at 11:05 PM, Miriam English <mim at miriam-english.org>
> wrote:
>
> > There is also the human-unfriendly problem of unmounting a drive before
> > disconnecting it. Everybody knows this must be done, but everybody has
> lost
> > data because of it. I have, on odd occasions, usually when over-tired,
> made
> > the mistake of pulling out the wrong USB connector or switching off the
> > wrong external drive.
> >
> > <rant>I'm certain the design of peripherals could have been made to take
> > account of the flawed nature of humans so that data was buffered in drive
> > electronics, and capacitors supplied just enough power to ensure pointers
> > were updated correctly. But USB, while having terrific advantages over
> > previous kinds of data connectors, was nevertheless a good example of an
> > over-complex, yet flawed, spec designed by committee. The main USB
> document
> > is more than 600 pages long!</rant>
>
> Horrible as USB is, I don't think it can take the entire blame here;
> the problem may be that the OS hasn't yet written back all the data to
> the drive at the stage the human pulls the drive out.  I guess it
> could be fixed at the OS software level, but at the cost of efficiency
> (more frequent drive accesses).
>
>
Could help to mount the drive as "dirsync" so at least the directory
structure is always updated as quickly as possible. Mounting it sync
generally degrades performance so much it's not really usable (although if
you only copy large files to and from it might not make a difference to
you, it's mostly when you use it interactively that async helps a lot,
because your writes go all over the place and are usually small).

Wim