Subject: RE: Netatalk and aliases
From: Jonathan Newman (jnewman@mudpup.com)
Date: Fri Jun 16 2000 - 17:13:45 EDT
> assuming the DID's are a certain format, and that directories under unix
> have a set inode (or id), couldn't an algorithm for creating a unique,
> repeateable AppleShare IP DID from the unix inode resolve this issue? the
Adrian Sun sent this message to the list in February. Note the comments
about DIDs and inodes. So there is an attempt at an alias fix in 38 and 39.
It doesn't work for everyone.
" Date: Mon, 14 Feb 2000 23:10:59 -0800
From: a sun <asun@cobalt.com>
To: netatalk-admins@umich.edu
Subject: test tarball available for omnis/fixed directory ids
hi all,
i just thought i should let everyone know that i've figured out why
omnis database doesn't like afpd. it turns out that omnis tries to
unlock non-existent locks. my locking code didn't handle that case. i
also went ahead and re-wrote the code a bit so that there's less
pointer twiddling, but i think it ends up being a wash in terms of
performance.
anyways, i've put a tarball for people to test at
<ftp://ftp.cobaltnet.com/pub/users/asun/testing/pre-asun2.1.4-38_test.tar.gz
>
it has the _test suffix because i currently have the codepage support
disabled. i'm in the midst of re-writing it to handle double-byte
codepages.
things i would like from testers:
1) does it compile under redhat 6.1 without complaints about
sendfile and libwrap?
2) does it preserve aliases correctly? i'm currently using a
24-bit hashed file/directory id field + 1 bit file/directory
identifier + 7 (4 major/3 minor also hashed) bits for the
device.
3) as a side-effect of what i did for #2, does sherlock run
more quickly?
4) does locking appear to work correctly? use your favorite
database program to test this.
-a
asun@cobalt.com"
Jon Newman
jnewman@mudpup.com
This archive was generated by hypermail 2b28 : Wed Jan 17 2001 - 14:31:04 EST