Re: svlogd -tt

[Gerrit Pape <pape@smarden.org>]

On Mon, Jul 26, 2004 at 08:39:39PM +0200, Henrik Heil wrote:
> >When making it a runtime option, there must be a decision whether svlogd
> >shall maintain two different sets of files in the log directory, or
> >filename conversion should be done.
> 
> Thinking about these alternatives i must admit that both seem to have a 
> drawback -- and that i underestimated the complexity of switching the 
> formats.
> 
> Different sets (if i understood that right) means that it is not 
> intuitive what the maximum number of old log files from the config 
> means. It cannot mean num=TAI+UTC because in this case svlogd could not 

Yes.  But even so this looks like the best way to do it if at all.
svlogd would maintain two different sets of log files (one set with
TAI-filenames, one with UTC-filenames) independently, according to its
configuration.  When changing the configuration, and "current" gets big
enough, svlogd adds it to the currently configured set, leaving the
other set of log files alone.  If desired, the log files could be
converted from one file name to the other manually using a separate tool
(this again raises the request for a tai64nlocal alike program), or
removed.

But I'm not convinced that it's worth the trouble to implement this or
similar.  There still is the idea of a log processor program (svlogp)
which could be also used to list the log files in a directory with human
readable names, or gives convenient access to the logs by other means.

> What do you think about a runtime option for UTC-filenames that only 
> comes into effect if there are no TAI-filenames in the log directory and 
> vice versa?

I'm not sure; svlogd at least should print a warning in this case, which
then ends up in the readproctitle log.  It's not that intuitive to
users, and susceptible to misunderstandings in my opinion.

Regards, Gerrit.