Re: Proposal: Nuke negative comments

On Tue, May 10, 2022, at 3:57 PM, Christoph M. Becker wrote:
> On 10.05.2022 at 20:21, Larry Garfield wrote:
>
>> There are currently tens of thousands of comments in the docs.  Most are of dubious value, but occasionally some are useful to fold into the doc page before deleting.  I and others do so periodically, but it's all ad hoc and not especially efficient.  One of the issues is that determining if a comment is of any use is a subjective human task.
>>
>> However, doc comments do have up/down votes on them.  While they are also of dubious value, I think it's a reasonable assumption to make that any comments that have negative totals are worthless and we can safely get rid of them without human intervention.  That would remove many thousands of comments from the docs, and give us a better sense of what pages have the highest number of potentially-useful comments worth trying to manually review, merge, prune, etc.
>>
>> This could be done manually through the admin UI, but as I said, "many thousands."  Would someone with server access be able to just nuke any negative-score comments from orbit?
>>
>> Anyone object, before we track down someone with the nuclear codes?
>
> I'm not quite sure about that.  For instance, most recently I found
> <https://www.php.net/manual/en/snmp.installation.php#115683> which has a
> -17 score, although I think that information is actually quite useful
> for those using snmp on Windows.  I have no idea why that got that much
> downvotes, though.  Is it because nobody cares about the folder layout?
>  Is it because Windows isn't an important platform?  Or doesn't it work
> (I didn't try it, but from looking at the code, it should work as
> described)?  Or is it because it only describes a part of the story?
>
> Of course, that is just a single example, and most downvoted user notes
> I've seen in the past were actually not helpful (at least I would agree
> on that), and I don't think that it would be too bad to delete a few
> potentially useful notes by accident, if we were doing a mass purge.
>
> Then again, the points raised by Dan are something to consider.  So
> yeah, I'm not quite sure about that.
>
> --
> Christoph M. Becker

I'm fine with some slightly stricter threshold for purging (-5, or -1 but also at least X months old, or whatever).  No strong preference here.

I agree it is likely that there will be a small number of false positives with this approach.  The value of the crowdsourced votes is, as noted, a dubious heuristic at best.  I have no idea how to estimate what that false positive will be or what "too high" would be.  If that particular comment is valuable, it should be merged into the page.  (I never use Windows, so I have no idea if it's valuable or misleading.)

To Dan's point, I am happy to discuss better options than this sledgehammer.  Mainly I'm just trying to improve the signal to noise ratio to make the janitorial work easier and more productive.  But I don't want a mythical future solution to block a simple and easy one now, either.  (Having a stricter one-time purge rule could still leave us sufficient data to test more nuanced tools in the future.)

I do not currently have server/git access myself, but I do have access to the management console on the web, and I've used it to nuke a bunch of low-end comments before.  But... there's a lot of them.

--Larry Garfield