-=> Mocking Richard to Kurt <=-
  (Mock, mOck, moCk, mocK)
 -=> My computer told me that Kurt Wismer said to Richard St. John
 
 KW> false... 400-500 is sufficient for an "in the wild" test only, the
 KW> viruses in the wild are not representative of the whole set or of the
 KW> actual protection a scanner can offer and virus bulletin's "in the
 KW> wild" test bed had only 97 viruses in it... and it tested only file
 KW> infectors... we all know boot infectors are more prevailent in the
 KW> wild than file infectors...
 RSJ> If that was the only testing they did on a scanner, then I would tend
 RSJ> to agree with you that their testing is not good. However they do test
 RSJ> the scanners with a bit more than 'in the wild' testing.
 
yes, but their test bed is too small... with just under 400 total
viruses, last years test was a joke...
 KW> i don't expect to see a huge improvement...
 RSJ> If the scanners would improve, then you might see some better testing
 RSJ> results.
the scanners are not the problem with those tests...
 RSJ> Virus Bulletin presents the numbers as they appear, they do not read
 RSJ> into the numbers or sway the results. All scanners are given the exact
 RSJ> same testing circumstances, same contamination, etc. Therefore the
 RSJ> numbers can be an indication of a scanners performance.
i disagree... with that unreasonably small test bed, the chances of a
statistically anomalous set of results is quite high...
 RSJ> It does sound like you would prefer listening to an anti-virus
 RSJ> company such as McAfee say,"Our virus scanner is the best because we
 RSJ> tested it ourselves against other products."? Rather than a 3rd party
 RSJ> company doing comparison testing.
untrue, i don't listen to those types of tests either... the best tests
i have ever seen done were by (at the time) third party
individuals/organizations... they only managed to use about half of all
the known viruses but that is orders of magnitude better than the virus
bulletin test last year... there are several requirements for a good
test and one of the most important ones is a large and sufficiently
random test bed... this is a requirement of any statistical analysis...
if i went out and surveyed five people on their opinion of something,
would you trust the results? i wouldn't - that small a sample can't hope
to be representative from a statistical point of view... the virus
bulletin test had the same problem (though to a somewhat lesser
degree)...
... beware of quantum ducks... Quark! Quark! Quark!...
~~~ TGWave v1.20 Beta-07+
--- Telegard 3.02/Gecho
---------------