On Wednesday, 22 August 2018 at 15:17:36 UTC, Jesse Phillips
wrote:
It is weird that you make loosing current and historical pull
requests is minor
It would be disruptive. However, work could resume rather quickly.
The disruption would be reduced if we had a periodic job set up
to mirror github pull requests. There is at least one export
tool, but it seems to get only the titles and not the comments or
patches.
compared to:
* Having all the data readily available for search engines to
have archived (today, not tomorrow).
* Having an established forum/newsgroup readily available to
handle the load of new questions.
I just don't see data retention and recovery for StackOverflow
to be a concern for making such a choice. Even if it did take
weeks or months to host the historical data, risk should be
weighed against possible benefit from visibility and growth
from heavily using StackOverflow.
And similarly, the choice of Github instead of a self-hosted
system is weighed against requiring people to sign up with a
private gitlab instance. Also similarly, the disruption would be
reduced if we had a periodic job set up to handle long-term
stackoverflow unavailability in advance.
I'm a little paranoid about centralized services like Github. I'd
prefer a federated service for source control / project
management, where you could easily fork projects from my server
to yours and send back pull requests. Then there would be no
extra cost for hosting your own vs using an existing instance.
I've been low-key thinking about making a federated github, one
where exporting your data is as simple as a `git clone; git
submodule update --init`. Probably nothing will come of it,
though.