0.92.6's bogotune much slower?

[Greg Louis]

On 20040905 (Sun) at 1055:55 +0200, Valient Gough wrote:

> Thanks for the description.  My bogotune run has just finished, so I can
> report that it took 5 days worth of CPU time to run.  Doesn't this seem
> excessive as the default?  Perhaps the flag should be to enable ESF
> search instead of disable it.

If you prefer that, it's a simple change at compile time.

> I just started a run with ESF disabled, and I notice that instead of
> using the ESF parameters that are in the configuration file, disabling
> ESF instead forces spesf and nsesf to 1.0.

It was assumed that if you don't want to tune ESF, it's because you
don't want to use ESF, which is what setting the parameters to 1
accomplishes.

> I've changed my local copy to use the existing ESF

It might be a valid option to let people ask bogotune "what are the
best min_dev, s and x with these specific ESF values?"  I confess I
hadn't thought of that (I would never do it myself, as cpu cycles are
of less value to me than optimal bogofilter settings).

(at least for the course scan):

In general, if you're going to tune ESF you should tune it, and if
you're not, you should maintain it unchanged.  Allowing it to vary
during the fine scan only is, I think, unlikely to do any better than
leaving the original values in place throughout.  If you disprove that
assumption of mine with a well-designed comparison, however, I will be
very interested to see your results!  It's true that the saving of time
would be attractive to many people if the outcome were nearly as good.

-- 
| G r e g  L o u i s         | gpg public key: 0x400B1AA86D9E3E64 |
|  http://www.bgl.nu/~glouis |   (on my website or any keyserver) |
|  http://wecanstopspam.org in signatures helps fight junk email. |