Glenn,

Good summary of the options.

If we are going to have a remote SNMP config system, then I vote for
B-2.

Just specifying a local filename to load (option B-1) still requires us
to get the file to the remote system somehow.

My intention is to have an app that takes a text file, convert it into
SNMP format and send it to the remote syslog. If the config was
accepted, we get an ACK response, otherwise we get an error response.

With Kiwi Syslog and WinSyslog, the config file is not created or edited
by the user. It is a file created by the app in INI or XML type format.
It is possible that the user could edit it, but not recommended. The
program properties are set from within the GUI app, then the config file
is created from the settings.

With Syslog-ng and SDSC etc, the config files are human
readable/editable. So a simple SNMP-SET type command line app that sends
a text config file would be a good choice here.

Rainer, does that that work for you?

Cheers

Andrew











-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Glenn Mansfield
Keeni
Sent: Sunday, 8 February 2004 7:21 p.m.
To: [EMAIL PROTECTED]
Subject: Re: SyslogMIB Issue-#4 // Issue-#2


Andrew Ross wrote:

> Having a minimum set of configuration options would probably be a
waste
> of time in my view as all implementations are so different.
>
> Since all implementations use (or can be made to use) a config file,
my
> vote is for a free form string variable. The format of the text
depends
> on the vendor. All we need to agree on is a standard cookie to
identify
> the vendor and version.
>
This makes sense.

If I have heard and understood correctly - we would love to have
configuration
options - BUT given the diversity of the syslog implementations we do
not have
one model for configuration that would allow us to develop a
one-size-fits-all
MIB.
However there is the trivial model - which uses the configuration file
to hide
all the implementation details. This model probably fits all
implementation.

That would mean  -
     A. We remove the configuration details
     B. We introduce a configuration file object. Here we have two
choices
        B-1. Just leave it at the fileName level of detail-  The
fileName is
             writable
             Manager will set the syslogConf file name and expect that
to
             be used. We have 1 MO in this case.
             Comment: This is simple. The manager aranges for the
syslogConf
             to be present (by non-SNMP means) and just tells the syslog
process
             to use the specific syslogConf file.
        B-2. Treat the file as a text file with one or more lines.
             Represent the file as a table: a row is a line.
             The semantics or syntax of the file is opaque to management
process.
             It can read the file and display it on the console (using
SNMP).
             For update of the file:  allow lines to be modified (using
SNMP-Set)

             Comment: This is much more complex. To do this nicely we
will effectively
             need a text file editor to be implemented using SNMP.

     C. At a later date, if there is interest and demand, we can have a
        BSD-Syslog configuration MIB where the parts removed from A can
be
        used.
        I liked what it was doing- it is pretty good and useful when I
want
        to remote-monitor (not control) the configuration of the syslog
process.

Let me have your comments.

Cheers

Glenn








Reply via email to