Ed Cashin wrote
> With "vbladed", I was probably influenced by Dan Bernstein, who says
> that a process should not daemonize itself when an external program
> could do it if and when necessary.  It certainly does tend to make the

Dan Bernstein also says it's OK to do things like put binaries in /var 
and not implement parts of well-known specifications you don't like for 
whatever reason. I take what that guy says with a grain of salt -- some 
of what he says is good sense, but other stuff is *way* off-the-mark.

> C code cleaner when the program is written for the general case and
> can be used either way.

well, sure, but adding some daemonize code in there doesn't make the 
source code *that* much dirtier. It really is a simple thing to do, and 
is much more convenient than 'eval sh -c "/blah/blah some stuff &"' if 
you want to fork something off into the background. adding a -x (or 
whatever) to keep it from forking or a -d to tell it to fork isn't that 
tough either.

> Regarding fadvise, I am not sure whether POSIX_FADV_RANDOM would be
> helpful, especially since the user might intend to do mostly
> sequential accesses.  It might be interesting to hear about cases
> where it does (and doesn't) help, though.

It might be good in the case where you're accessing a database located 
on AoE storage. But then, in my experience, fiddling with the I/O 
scheduler has a much greater performance impact (in the positive) than 
the posix call would. But as always, results may vary and I'd be 
interested to see what the numbers do when this is turned on.

regards,



-- 
Kelsey Hudson
Sr. Systems Administrator, DrJays.com
9180 Camino Santa Fe, San Diego, CA 92121
888.437.5297x134 (desk)    619.852.6374 (cell)


------------------------------------------------------------------------------
_______________________________________________
Aoetools-discuss mailing list
Aoetools-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/aoetools-discuss

Reply via email to