if you take this logic and continue with it just a bit, you will realize that the picture+text method discussed has no advantages over pure textual method.Kevin Clark wrote:Kellanved wrote: Nope, this would work for individual forums, but not for a standard distribution.
The reason is quite simple: all bots would "know" the answers for all the default questions about a week after its release.
But if it was configurable so admins could change the images to their own, in a very simple manner, each forum would have a different 'answer'. You would only have to make a picture in MSpaint (words, images, a code, anything) and upload it with something else to configure what the accepted answer is.
I understand your point that the bots would learn the defaults very quickly but if it was simple for the end users to change it then it becomes much more difficult and everyone images would be different.
disadvantages:
-- requires some graphical tool(s) to create the image(s)
-- about 10 times more work to prepare a new question (once you have the tools)
-- not usable for the visually impaired
-- no straightforward way to log which question each user have answered upon registration (logging is useful because once you discover a spammer you can replace the question)
-- takes more space on the page than strictly textual question (not important on registration, but is important if you want to use the challenge for guest-posting)
-- can't be used by text-only user-agents
advantages of the "picture challenge" over text only challenge:
-- none.
of course, as you and many have noted, the picture/text or text challenge method becomes completely ineffective if the standard installation already comes with such a challenge.
however, if standard installation doesn't come with any pre-made challenge, the text-only method is beautiful in it's simplicity, and i believe can be very effective.
the only real problem i see is that if/when the board operator creates a challenge which is too difficult, such as asking a question whose answer is ambiguous, or may not be known to all potential users, or worse yet, if the person creating the challenge have keyed in a wrong answer (either plain wrong or misspelled), nobody will be able to register, and the operator will never know the reason, because in order to alert the operator one has to register first...
(curiously, exactly the same issue exists with non-human-solvable captcha, or with erroneous picture+text challenge...)
some anti-spam measures only make sense if they are part of the core package. most notably, marking each user-supplied URL with rel="nofollew", which was discussed several times.
other anti-spam measures will work just as well, or maybe even better, if applied as MODs, because it will splinter the board population, thus making the job of the bot-creators more difficult.
luckily, both methods discussed (picture/text challenge and pure text challenge) belong to the 2nd group, do i don't see this issue as critical for the RC.
personally, i feel that the devs made a wrong call when they decided not to add the rel="nofollow" to user-supplied links. i believe some other packages (blogs?) do use this, and while it will not prevent spamming, it's a small help in this "global war on spam", and has no detrimental effect whatsoever on the board, being 100% transparent to the admins AND users.
