The question of course is: how is it performing?   Are the current rules useful
or do they need work?

I'm not sure that things like "image size" is useful long term, or possibly even
the short term.

I think they look pretty good:

http://ruleqa.spamassassin.org/?daterev=20070108-r493989-n&rule=&srcpath=imageinfo.cf&g=Change

0.00000   6.9502   0.0071   0.999    0.96    1.00  DC_GIF_UNO_LARGO   
0.00000   7.2222   0.0142   0.998    0.95   (n/a)  __GIF_AREA_180K   
0.00000   5.6323   0.0089   0.998    0.95    1.00  DC_IMAGE_SPAM_HTML   
0.00000   4.6297   0.0089   0.998    0.93    1.00  DC_IMAGE_SPAM_TEXT   
0.00000  17.2178   0.1579   0.991    0.88   (n/a)  __GIF_ATTACH_1   
0.00000   9.5007   0.1313   0.986    0.87    1.00  DC_IMG_HTML_RATIO   
0.00000   8.8861   0.1579   0.983    0.85    1.00  DC_IMG_TEXT_RATIO   
0.00000  22.8236   0.4915   0.979    0.76   (n/a)  __HTML_IMG_ONLY   
0.00000   0.6090   0.0160   0.974    0.75   (n/a)  __PNG_ATTACH_1   
0.00000   2.5057   0.2271   0.917    0.74   (n/a)  __JPEG_ATTACH_1   
0.00000   0.6236   0.1065   0.854    0.68   (n/a)  __GIF_ATTACH_2P   
0.00000   0.2720   0.0071   0.975    0.66    1.00  DC_GIF_MULTI_LARGO   
0.00000   0.2139   0.0035   0.984    0.64   (n/a)  __PNG_AREA_180K   
0.00000   0.2132   0.0035   0.984    0.63    1.00  DC_PNG_UNO_LARGO   
0.00000   0.2118   0.0923   0.697    0.56    0.00  T_DC_IMAGE001_GIF   
0.00000   0.1119   0.0426   0.724    0.54   (n/a)  __JPEG_ATTACH_2P   
0.00000   0.0014   0.0000   1.000    0.49   (n/a)  __PNG_ATTACH_2P   
0.00000   0.0007   0.0000   1.000    0.49    0.01  T_DC_PNG_MULTI_LARGO   
0.00000   0.0000   0.0000   0.500    0.49    0.01  T_DC_GIF_264_127 

DC_GIF_UNO_LARGO in particular works great for me -- hits 19% of my spam!

actually, looking at the overlap data:

http://ruleqa.spamassassin.org/20070107-r493683-n/DC_GIF_UNO_LARGO/detail#tOVERLAP_new

I initially thought there was a bug in that it claimed that no other rules had
an overlap ratio higher than 0%. ;)  Turns out it's true; I checked the logs. 
There are very few rules that overlap much with DC_GIF_UNO_LARGO, and seeing as
that's a very good rule (hit-rate-wise), too, I think it's valuable; it adds
diversity to the hits we get (overlaps are generally a bad thing).   

I'd be +1 to putting the plugin into core, given that.

Although running, ImageInfo is not catching spam at all for me. But mainly 
probably due to my inability to weigth the different options, due to lack of 
understanding of them due to lack of documentation.
Plug in seems somehow promising but we "normal" users not able to read the code 
need some guidance on its use.

Anyhow, image spam is on the rise, and both ImageInfo and FuzzyOcr give what 
they give ... Probably something like this http://www.nabble.com/forum/
ViewPost.jtp?post=5924483&framed=y sounds more promising ... And, for those of 
us behind a firewall, razor2, DCC and the like are a no way.

Unfortunately I am not able to code otherwise would volunteer to code the 
spectral analysis plug in.

CT

just checking licensing:

------------------------------------------------------------------------
r430232 | felicity | 2006-08-10 04:55:31 +0100 (Thu, 10 Aug 2006) | 1 line

w/ permission, grabbed the latest imageinfo plugin and ruleset for testing purposes
------------------------------------------------------------------------
...
------------------------------------------------------------------------
r428576 | felicity | 2006-08-04 00:55:38 +0100 (Fri, 04 Aug 2006) | 1 line

test plugin and rules from Dallas L. Engelken ala the dev list
------------------------------------------------------------------------

since it's in for "testing" only, worth checking -- Dallas, are you happy to put
this into the main ruleset?

no reply.  it's looking like a "no" right now...

(In reply to comment #6)
> no reply.  it's looking like a "no" right now...

It was donated to us, I just have left it in my sandbox area, so if you want to
move it in as a standard plugin, go for it.

ok, cool.  I'll do that then.

ok, applied.