Comments on this blog hasn't been implemented yet, however I'm close to finish the design of how they work. Long story short, the comments and tracebacks will not change anything on the server side, but will be sent to a queue for moderation/acceptation that can be processed locally on the machine where I do the actual posts.
The benefit for this approch are mainly:
- server-side data stays read-only from the webserver point of view.
- all comments/tracebacks are to be approved before actually posted: net effect on reducing spams.
- duplicated and OT comments can be filtered out.
- comments related to post update (something missing in the post for example) doesn't have to be posted, but just acted on.
As a simple prototype, my initial queue design is going to be on top of SMTP and email. Each time a comment or traceback is receive, instead of modifying a file or a database on the server side, a simple email will be sent to a mailbox after basic checked where issued (spam checking mostly).
As this point i can just process the queue of comments with favorite $MUA. The acceptation process is simply piping the email through a special adding comment binary, that will add the comment to my local data. All this just a simple $MUA macro that does the piping with a simple and swift keybinding. At this point the comment is not visible but will be as soon as I push the data back to the server.
Eventually, new queueing system can be developped on top of XMPP, AMQP. Also the queue processing could be done by a simple bot when the data doesn't have to be moderated.
