It is true that qmail doesn't do anything with the inode->filename
mapping after it is made, besides have unique file names, and that
replacing that algorithm (which certainly succeeds in providing
insight into How The File System Works) with a different algorithm
that also guarantees uniqueness would break nothing?




-------- Original Message --------
Date: Thu, 13 Jul 2000 21:38:42 -0700 (PDT)
From: dean gaudet <[EMAIL PROTECTED]>
cc:"[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
Subject: (reiserfs) Re: Jedi's qmail reiserfs integration status report

On Fri, 14 Jul 2000, Hans Reiser wrote:

> I strongly believe that it is not deep programming to find a portable version of
> the algorithm.

yeah.

i'm not sure how few bits you guys want... but there's the code i put into
apache 1.3 for mod_unique_id which generates a 112-bit unique id subject
to a bunch of completely reasonable constraints.  see
<http://www.apache.org/docs/mod/mod_unique_id.html> for documentation.  
feel free to snarf the code under any license -- it's essentially
unchanged since when i put it in there.

it's basically a <timestamp, local_ip_address, pid, counter> tuple.

we replaced qmail's filename generation with this at cp -- mostly because
we're still living with solaris 2.6 and it doesn't cache filenames longer
than 31 characters; the shorter filenames helped a bunch.

-dean

Reply via email to