|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
Subject: Re: SCSI, U2W in particular?, and poor performance...
From: Wietse Venema (wietse
porcupine.org)Date: Sun Mar 05 2000 - 15:53:54 CST
- Next message: Rich Graves: "Re: SCSI, U2W in particular?, and poor performance..."
- Previous message: Liviu Daia: "Re: Is this possible with postfix?"
- Maybe in reply to: Justin Robertson: "SCSI, U2W in particular?, and poor performance..."
- Next in thread: Peter Wemm: "Re: SCSI, U2W in particular?, and poor performance..."
- Maybe reply: Wietse Venema: "Re: SCSI, U2W in particular?, and poor performance..."
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Craig Sanders:
> On Sun, Mar 05, 2000 at 01:56:29PM -0500, Wietse Venema wrote:
> > (*) for that, Linus recommends that the application open()s the
> > directory and fsync()s it. Right. When pigs fly. Fortunately we still
> > have a choice of operating systems.
>
> what exactly is the problem with doing that if:
>
> a) it can be wrapped in "#ifdef linux" so it doesn't affect compilation on
> other systems, and
> b) doing so will safely speed up postfix on linux?
So would migrating to a better file system, and that would have
the benefit of avoiding unnecessary code that jumps hoops whenever
Postfix creates, renames or removes a queue file, bounce/defer log
file, mailbox file or maildir file.
> is it just an aesthetic distaste for doing something "ugly" or is there a
> good technical reason for not doing it?
I have something against unnecessary code that can't be tested.
Every line of code is a potential bug. Microsoft has a goal of 4
bugs per 1000 lines; my goal is fewer bugs than that. Code that
isn't written doesn't have bugs. That's how I avoided a lot of bugs
with Postfix.
> ps: yes, using reiserfs would be better - but it's not a standard part
> of the kernel yet and many people won't (or don't know how to) use
> experimental patches.
Reiserfs still needs to prove itself.
Wietse
- Next message: Rich Graves: "Re: SCSI, U2W in particular?, and poor performance..."
- Previous message: Liviu Daia: "Re: Is this possible with postfix?"
- Maybe in reply to: Justin Robertson: "SCSI, U2W in particular?, and poor performance..."
- Next in thread: Peter Wemm: "Re: SCSI, U2W in particular?, and poor performance..."
- Maybe reply: Wietse Venema: "Re: SCSI, U2W in particular?, and poor performance..."
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]