My post “Bug report: tkinter.scrolledtext isn’t applying OS copy/paste standards” (I can’t post 3 links) is currently hidden (as spam?). It is a real question, can you please unhide it?
I had the same problem for https://discuss.python.org/t/non-optimal-bz2-reading-speed/6869 (solved in the meantime), you can see it’s a real technical discussion, with real interactions.
Yeah, that doesn’t look at all like spam. Rather, it is the start of a very interesting discussion.
Unfortunately this thread is currently inaccessible to ordinary users, such as myself, because it was marked as spam. It displays this unfortunately familiar message:
Can you maybe whitelist my account so that I don’t get automatically flagged a priori in the future? Also if the “3 links are not allowed” (sometimes it’s even blocked with 1 URL) limit can be removed for my account, it would be useful - linking to precise posts on bugs.python.org or Github is useful.
Yes, a hard-coded solution might be needed for this case. But, could someone please educate us, in simple terms, regarding how the Akismet algorithm works? (Hi there, folks at Akismet, if that link got your attention.) It would be ironic if that link got my reply hidden as spam.
At the very least, an explanation might help users avoid having their posts flagged as spam. However, it would be a shame for that avoidance to degrade the quality of their posts.
Does Akismet use machine learning in their algorithm? If so, they could maintain a database for each site. For instance, they would have one for Python Software Foundation Discourse. For a finer-grained consideration, they could store information about each user of each site.
Each time they hide a post as spam, they would keep a history of the disposition of that occurrence. This would include what criteria caused it to get flagged and whether a Moderator verified it as spam, or unflagged it. Then the algorithm could learn that a specific user on this site, such as @josephernest, tends to include links in a post, and could increase the limit on the number of allowed links for that user. Other criteria could also get refined, as appropriate.
If a post that was not marked by Akismet as spam gets flagged by a user, the history of that occurrence would also go to the database for the edification of the algorithm.
When I ran afoul of some misconfigured spam filtering on this forum
last year, it also became apparent another way in which “mailing
list mode” users are implicitly second-class citizens. The Discourse
software seems to calculate your account’s reputation not by the
number and quality of the posts you’ve made, but by things like how
much time your browser spends open on different forum pages
“reading” and how many “likes” you’ve made. After a year of fairly
active participation here, my it still listed my reputation as
nonexistent because I don’t interact with the site through a Web
browser as though it were a social media platform.