Re: corba config caching

[quatt001 bama ua edu]

> The present solution definitely has its problems. It has traveled the
> pendulum between race conditions and suboptimal performance several times,
> and is not 100% "nice".
> 
> Further investigation would help shed some light on exactly how the
> performance would compare. I'm not saying either way is the right
> solution, just that there's no conclusive information in either direction.

Seeing this thread got me thinking of something I had half-coded
before.  Has anyone here ever put any thought into a sort of standard
config-file format?  As in, any configuration file that contains
user-configurable stuff would be in a standard format which is
hand-editable, but of course also parsable.  A side effect of this is
that it would make creating a generic config file editor possible.

Of course, if gnome config files are already supposed to be in a
standard format, then forget I ever said anything.  I'm assuming it's
application-specific, though.

I did some work on something like this a while back but never finished
it (yet).  The Specs library (that's what I called it, but I may change
that eventually since I found out that RPM uses something called spec
files as well) contains a set of functions that let you get and set
configuration values without having to actually touch the config file
itself manually (similar to the way Windows has always done .ini files,
but with a few extensions).  The config files are supposed to be
self-documenting, meaning that there should be descriptions of each
field in the file within the file itself (this makes generic
configuration editors actually useful!)

Anyway, I thought something like this might be useful to gnome,
especially since it would reduce the time coders have to spend rolling
their own config-file parsers (and because standard file formats are
always fun :)  Is anyone interested in this?  I can supply more
information if prodded - this was just a quick sketch.

To tie into the original thread - if it were decided that using an
intermediate binary form to speed access would be a good idea, the
confi-file library could handle that transparently, as well.

-- 
****
Stephen Quattlebaum
quatt001@bama.ua.edu

Death multitasks.  Someday, your timeslice will be up.

The opinions expressed by me are not necessarily the opinions of the
University of Alabama or it's faculty/administration (but they should be).