Run mollom on comments already posted

It might be nice for people installing Mollom to provide a 'Send to Mollom for textual analysis' comment/node operation where the user could also select which operation/action could be performed on the ham/spam results.

The problem with that is that we'd have to figure out all kind of access permissions and totally weird + complex stuff, which we're actually trying to remove in #717212: Remove "report to Mollom" links and integrate with entity delete confirmation forms instead currently.

For me, this feature request sounds like "nice to have", but then again, also "so why didn't you install it from the beginning, dude?" -- effectively, won't fix.

Agreed on the reasoning, but still, not very easy to implement.

From a technical perspective, this is almost the same as #245364: Provide API to programmatically validate content (integration with Trackback module)

EDIT: sorry, linked to the wrong issue. Fixed.

Hehe, im in a similar situation....i will certainly be installing it from the start on my next site.

Some googling got me here.

This feature would be useful for companies that inherit a client with a site that never had spam protection and need to clean it up. It would be nice to be able to run a report on the content/comments in question and get a ham/spam/unsure response.

For now I figure I will just query the nodes I have in question and manually pass them to the mollom api.

I had the same problem and wrote a quick Drupal 6 Batch API implementation (https://drupal.org/sandbox/rocketeerbkw/2149163).

I was able to delete ~25k comments. I'm pretty sure I'm duplicating some functionality built into mollom module, but I haven't been able to parse the whole thing yet. I'm also doing a "best effort" at mollom API errors, but the documentation seems lacking in that area. If there are rate limiting responses I'm not seeing them.

I would advise strongly against that type of batch submit from the client-side. You're likely to get poor spam classifications and could even negatively affect Mollom as a whole.

Essentially, by submitting comments in rapid succession, all of the authors look like spammers which can lead to misclassifications. Because Mollom learns from its classifications, it actually leads the whole system being trained incorrectly.

If this was to be accomplished, it would have to be something like what Dave Reid suggested so that Mollom knows a batch submission is in process and can react accordingly. This is something that the Mollom engineering team continues to consider and I'll pass along that there is continued support in the community for a process of this type.