Why not syslog?

• Messages sorted by:[ date ][ thread ][ subject ][ author ]

Russell Nelson wrote:
> This is not a flame; it's not even particularly related to mgetty.  Could
> someone tell me why authors of software invent their own logging scheme
> instead of using syslog?  Taylor UUCP does it, mgetty does it, modgetty
> does it.  Enough people do it that there must be some drawback to using
> syslog that I'm not aware of.

Yes, there is a major drawback: there are lots of systems that do not have
syslog. (Actually, I started developing mgetty on such a beast...). There
may even be systems that do not even have sockets to provide your own
syslogd with mgetty (not that I would be keen on writing one). Further,
syslog() cannot do all that mgetty's lprintf()/lputs()/lputc() functions
can do, so you would have to have a wrapper anyway.

Another point against syslog is: it is quite flexible, but not good enough
- there may be some reasons not to have mgetty output intermixed with
other system logging stuff, and the fixed set of facilities make it
difficult in complicated environments to find one that only mgetty uses...

Besides, mgetty *can* log to syslog, but only "higher priority" messages
are actually sent (that is, L_AUDIT, L_ERROR and L_FATAL), because I do
not want to load the machine by passing lots of debug stuff over sockets
that will be discarded anyway on the other end...

We've discussed using syslogd as the sole method for mgetty logging, and
there were some loud voices against it, namely portability and system
load issues. That's why I kept the old logging scheme and extended it with
optional syslog logging.

gert


-- 
Ok, Ihr habt gewonnen, hier ist eine neue signature...

Gert Doering - Munich, Germany                             gert@greenie.muc.de
fax: +49-89-3243328                         gert.doering@physik.tu-muenchen.de