Given that deleted content is licensed CC just as the regular content is, it exists in the current SO database, and it is accessible to over 400 (10k+ rep) users, it should also be exported as part of the regular data dump.

Further, it would give us someplace to send those annoyed with content deletion so they can retrieve what they feel they've lost.

Lastly, it would give statistics junkies another bit of objective information so we can better handle questions on meta regarding post deletion.

At minimum, if there are compelling reasons to keep the post body itself out of the data dump, then the post data can be scrubbed, while leaving the post id, title, ownerid, dates/times/tags/etc intact so that 10k users can go retrieve information when people come here asking about a deleted post without having the ID.

  • 2
    No - deleted posts occasionally include content where there was not a meeting of the minds for licensing or where something inappropriate for public distribution occurred. A visitor to stackoverflow does not leave their privacy or other rights at the door. There isn't enough notice for that in a link or logo at the bottom of the page.
    – Paul
    Jul 15, 2010 at 8:31
  • 2
    @Paul Those can be purged, super-deleted. Those don't show-up for 10k+ users either.
    – Mast
    Oct 3, 2019 at 6:32

4 Answers 4


Not unless there is an easy method to also disassociate oneself from the deleted material in accordance with the retained moral rights of the author under CC-Wiki.

The most reasonable assumption for deleted material (that remains deleted) is that the author wishes to remove his or her association with it. The CC-Wiki license retains to the author the right to have their associations with the material removed upon request. If the material remains deleted (sometimes an author may delete in order to edit and improve, then undelete), then either the material should not be included in the data dump, an easy mechanism should be made available to allow the author to disassociate him/herself from the material, or the material should be automatically anonymized upon inclusion. Once the dump has been made available it is no longer possible to remove the association in any copy that has already been distributed.

  • 3
    But the association exists in the database, and over 400 people can see it - therefore SO is breaking the CC-wiki license? At any rate, this can be taken care of by assigning the content to the community user when it is deleted by the owner - there are still many things that are deleted through other processes.
    – Pollyanna
    Dec 21, 2009 at 17:06
  • 3
    Or, to be more explicit - keeping the posts in the data dump is no more breaking the license than the system already is. So keeping them out of the data dump due to licensing restrictions is not a valid complaint.
    – Pollyanna
    Dec 21, 2009 at 17:10
  • 7
    The fact that the current system doesn't make it easily possible to disassociate yourself from deleted material, doesn't negate the real responsibility to safeguard the author's moral rights. The data dump crosses a boundary, IMO. It makes the material publicly available to everyone (not just "community moderators") and, once distributed, removes the ability of the SO team to enforce the author's moral rights. I think there is an obligation to respect these rights prior to the production of the data dump.
    – tvanfosson
    Dec 21, 2009 at 17:18
  • 4
    Also, I would argue that it ought to be easier to remove your association from the material on the live site as well. The fact that it isn't or that a person don't choose to remove their association knowing that it is available to 400+ people doesn't imply that they wouldn't if they knew that it would be distributed publicly without any restrictions.
    – tvanfosson
    Dec 21, 2009 at 17:20
  • It should permit to dissociate yourself, but the material should be available I think (there are 2 feature-requests here then)
    – juan
    Dec 21, 2009 at 18:22
  • 2
    I'm not going to argue licensing and legality here - you are severely mis-interpreting the license. But that doesn't matter. The data is available via CC-wiki to over 400 people. It doesn't matter if it's available a little bit, a lot, publicly, or to a restricted subset (since they are allowed to redistribute, it's the same exact license). It's all the same distribution. It doesn't cross any line - we're already way over any line you might suppose exists (it doesn't, btw). So your argument that it shouldn't be included in the data dump is comepletely and utterly without merit.
    – Pollyanna
    Dec 21, 2009 at 18:35
  • 2
    Re dissocating ownership of a post, this was requested here: meta.stackexchange.com/questions/31741/…
    – Ether
    Dec 21, 2009 at 18:49
  • Now I agree with Pollyanna again, it doesn't matter if it's associated
    – juan
    Dec 21, 2009 at 19:41
  • 1
    You miss my point. My choice to exercise my right to disassociate my name from the material may differ based on the use to which it is put. Nothing in CC-Wiki removes my right to make this choice. As it is deleted material is not included in the derivative work (the dump). The fact that I don't choose to exercise my rights with respect to the live site does not remove my right to exercise it with respect to the data dump. Thus either the data dump should handle it differently or it should be much easier to remove association from the material. Once in the dump there is no more control.
    – tvanfosson
    Dec 21, 2009 at 19:51
  • But what about the not-deleted content? Can't you exercise your right on that too? Isn't it already out there?
    – juan
    Dec 21, 2009 at 19:56
  • @Juan Manuel -- see my first comment. The answer is yes, but deletion is an obvious indicator that the author probably doesn't want to be associated with the content.
    – tvanfosson
    Dec 21, 2009 at 21:01
  • 1
    "My choice to exercise my right to disassociate my name from the material may differ based on the use to which it is put." - Unless you dual-licensed your posts when you made them then you cannot make that distinction. Further, you are contributing your posts to SO without reservation - when you make a post you do not license it to SO under CC-Wiki. The CC-Wiki license concerns material that SO distributes - as far as the law is concerned, once you hit post you have given them a full, non exclusive right to publish your content in any form and under any license they so choose.
    – Pollyanna
    Dec 21, 2009 at 21:39
  • Rather than say dual licensed, I should have said, "made an exclusion or distinction based on method of dissemination"
    – Pollyanna
    Dec 21, 2009 at 21:42
  • I don't think so. The data dump is a transformation of the work to which I contributed my content -- the SO web site. As such it is also licensed under CC-Wiki and must have a license that is the same or similar to the one under which I licensed my content. This license does not abrogate my moral rights. I have the same rights under the data dump as I do in SO and need not exercise them in the same way.
    – tvanfosson
    Dec 21, 2009 at 22:16
  • 2
    @Pollyanna -- @Jeff Atwood apparently disagrees with you: meta.stackoverflow.com/questions/13976/…
    – tvanfosson
    Dec 22, 2009 at 15:46

I've thought about this for a while, but I still don't see the point. Presumably, the posts were deleted for a reason: answers that were incorrect, unhelpful, or spam; questions that were off-topic or otherwise inappropriate... And so they were removed from the site (or at least, their visibility was cut drastically, since as you note a subset of users can still view them if necessary).

...and that's good. It reduces noise on the site. If a deleted post - or an older revision of an edited post, for that matter - is copied and hosted elsewhere, or cached in a search engine, or remembered by someone who read it, that's none of our concern. But pushing them into the data dump seems counter-productive: now it's no longer a "snapshot view" of the site as seen by an average user, it's... something else. Every client has to take deleted posts into account, figure out whether to display them or not, etc. Queries would have to be modified.

And for what? Why struggle to retain information that the site itself considers useless? Why not ask for a full revision history as well then. Shucks, even that would be more useful - you could have some fun with that, track editing contributions or something. Just to placate users who didn't care enough about the crap they posted to save a link, or even notice that it'd been closed? Ugh... If they really want it that badly, they can ask a moderator for help.

  • 1
    Meta information about deleted posts would be super useful though. Think about the history of closed, then deleted questions - it would become accessible to data mining.
    – Sklivvz
    Oct 3, 2019 at 6:38
  • 2
    The problem with having an 11-year history on a site is... Stuff like this doesn't age very well. A year in, there just wasn't that much deleted stuff. There were two mods, and Jeff. There were a countable number of people with delete privileges. And, of course, self-deletions which would still be problematic. We didn't have auto-deletions, or review. I should note that we did add some metadata on deleted posts a few years back: see the PostsWithDeleted table.
    – Shog9
    Oct 3, 2019 at 13:57

For data mining purposes it might be sufficient to retain

  • Creation and editing history (times and responsible parties)
  • ID
  • Parent ID (for answers and comments)
  • Type of deleting

but not the actual content.

However, this does not address issues like My question was deleted, can I get the infomation/answers from it?.


Deleted posts can be extremely useful for researchers as well. I have been struggling to predict the quality of the questions before posted on the site using NLP and Deep Learning. Other researchers work similar topics. Deleted posts provide negative examples. Although there are negative examples also in dumps their cardinality is much less than the actual negative posts therefore the distribution of the data is distorted.

Analyzing the content of the deleted post could help to develop tools, which might assist users to test their questions against the policy of the site before posting. I agree with the community that these data are noise for most users, but for research, they are valuable. For this purpose, these posts might be provided in anonymized manner.

  • 2
    If you have a clear research goal that is also beneficial to SE in some way, you can contact them to obtain a dump including deleted posts. You do need to sign an NDA in that case.
    – rene
    Oct 3, 2019 at 6:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .