Mon 13 Feb 2006 10:28:27 AM UTC, original submission:
We should provide users a mean to mark comments (or even items) as spam.
Catching posted spams:
--------------------------------
There would be an extra button around comments.
There was many ways to achieve this, the best one will be the one that does not require manual assistance.
We could allow only tracker manager to mark a comment as spam but I would lower efficiency of the thing.
We can allow any users to mark a comment as spam. Any users. That's what I have in mind. Each users will be able to mark whatever item (comments, items themselves) as spam.
Each comment or item will get a spamscore. If the spamscore of 5 is reached, the item is said to be spam.
A normal user non-member marking a comment increase its score of 1. It would take 5 non-members to mark a comment as spam.
A project member would increase it by 3. It would take 2 members to mark it as spam.
A tracker manager or project admin would increase it by 5. So only 1 would be enough to mark it as spam
Spam marking should probably not implies any notifications, to avoid it making more fussy than needed.
Hiding spams:
-------------------
What should we do with spams?
One way to go would be to hide the comment marked as spam and show it on click, with a javascript.
Nice in theory, but no good in reality: hyperlinks in the spam would still be in the page, making it available to web crawler that are the target of many spams, spam would still be there for people without javascript (hence breaking the rule of having a site that works without javascript)
I guess the proper way to go is to hide the comment, just providing a simple link to the message.
Stopping spammers:
----------------------------
I guess that whenever an authenfied user message is marked as spam, it should mean something about the user.
Whenever someone mark an item as spam, the user should get it own spam score increasing. And that would be used to set its default comments score.
There are high risks of censorship here. So we have to restrict this.
X post a comment. Y that is admin or manager of the project mark the comment as spam. X spamscore increase of 1.
Next comment X post get by default a score of 1. Making it easier to mark it as spam.
The limits are: only admin or manager spam mark would affect user spamscore; user spamscore can get higher than 4 (Otherwise it's messages would automatically be ignored, that is censorship).
On the admin front page, there could be an highlighted showing number of users that got a score of 4, allowing the admin to fastly deactivate their account after checking incriminated comments. I'm not sure about this: this would increase admin load and would be useless if it still easy to create new accounts.
Accounts should be deactivated but not trashed: history would be kept and... that matters, we would still have on file the email address used to create the account.
Catching spams when posted:
---------------------------------------
Maybe we could think of having savane interact with razor2 and spamd, especially for anonymous post, that would simply refuse the post when a high spamscore is detected an ban the IP for a short time.
|