tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Removing softdep

"Martin J. Laubach" <> wrote:
>   It's probably my age showing, but I'm wary of throwing out something
> that surely has its share of rough edges and warts but works for many
> people with something new solely on grounds of a promised brighter future.
>   Bring the new thing in, let it be tested, improved, broken, mended,
> abused and when it has proven to hold its promises, THEN AND ONLY THEN
> start talking about removing softdeps.
>   NetBSD should not fall for the buzzword technology du jour charm and
> premature removolation.

Well, softdep always broke SQL databases for me after unexpected power
outages, until I have stopped using it... :)

Anyway, I think the essential point is code complexity, which leads to the
bugs. UFS/FFS is already complicated code, and if removal of softdeps, and
addition of journaling will keep the equal functionality (does not matter if
it will be slightly better or worse), and will simplify the code - it is
definitely worth.

Another very important thing - NetBSD prior 5.0 used kernel-lock, which made
locking issues in many components hidden. Some of them fixable, some of them

Thus, I think removal of softdep would be another good decision based on
software engineering.

Best regards,

Home | Main Index | Thread Index | Old Index