probertm at acm.org
Sat Aug 27 01:38:02 EDT 2005
On Fri, 2005-08-26 at 17:36 +0300, George Moschovitis wrote:
> After a quick look I do not understand why this is better though. (And
> in fact I don't understand Zed's problem too).
I'll see if I can do it justice. Og is one part of my application,
let's say a data collector is another. The connection info comes from
the database via Og, into the collector, which fires off a threaded
collection. The data comes back, is parsed, trimmed, reduced, and
stored in the database.
So, there is a logger for Og, a different one for the collector and a
third for the parser and analyzer. Why? Cause different people are
interested in different things, with the Loggers being logically at
different levels. For example, if I have a communications problem
collecting data, why have a whole bunch of database info gumming up the
logs? Easier to have different logs than to have one and have to filter
it all the time.
The current design of the Glue logger
a. makes it a singleton, so there is only one log possible
b. changes the format, and I need the timestamp info for the comms
Unless I am mistaken, which I could well be. A GlueLogger "fixes" both
of these issues at minimum cost and no impact to third party libraries.
Just my $0.02.
More information about the Nitro-general