Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
As many of you know, the nLab is currently hosted on a server at CMU, financed by a grant for HoTT there. I am not sure how long there is to go on that grant, but I am not sure it is longer than a couple of years. I would suggest that it might be a good idea try to begin applying for some further grants fairly soon. How shall we go about this?
I am also a little bit worried about whether there are any backups of the nLab or nForum currently if the server at CMU is hit by lightning. To safeguard years of work, we need some good safety net in place. One option, if we have the money, is to use cloud storage, e.g. Amazon’s. Their set up is such that it is basically impossible that the data could be lost. It is pay per use, and a priori quite cheap, but if we host the actual database there, since it is up 24/7, costs might not be completely trivial. We could perhaps use the cloud only for backup, say once a day, and maybe use an S3 bucket rather than a database, which would be a lot cheaper, but still not completely negligible.
Another option is to try to initiate some kind of ’goodwill system’ amongst a collection of universities across the world to call our server once a day, say, to get backups of the data. We could have a minimal web application hosted at CMU which monitors these, so we know they are taking place.
It’s great to see that your are thinking about this!
Regarding the hosting, I imagine the hosting at CMU cannot be all that expensive (?) and the end of one grant is often the beginning of the next. So here we should talk with Steve Awodey, to hear what long-term perspective he sees of the present arrangement.
Regarding the backups: Adeel had set up a github mirror of the nLab at github.com/ncatlab/nlab-content-html. For details see his message here. Might that be sufficient?
Good idea to speak with Steve Awodey! Would you like to send him an email, or shall I? I have no idea how much the hosting costs at CMU. But running something like the nLab in the cloud with a reasonably powerful server and database would definitely have some cost; my guess (which I’ve not thought about carefully, only comparing with other cases I am familiar with) would be perhaps a thousand US dollars a year, maybe a bit less, maybe a bit more.
Ah, yes, those backups are sufficient for the nLab for the moment. I actually thought that github would not welcome this kind of thing, but since it obviously is working, it is fine for now. It would be non-trivial to get the nLab back up from those HTML files that are backed up, but the data is at least there in some form, which is the most important thing. The nForum is not I think backed up in this way, we should do that as well. I will look into it when I get the chance.
Okay, I’ll prepare a message to Steve.
Regarding backups: Is it just html or is it the Instiki source that is being referred to in Adeel’s message:
To save server resources you can download a “bare” version of the repository with
git clone --bare nlab-git@saunders.phil.cmu.edu:nlab-content
and then run
git clone nlab-content.git nlab-content
rm nlab-content.git
cd nlab-content
git remote set-url origin nlab-git@saunders.phil.cmu.edu:nlab-content
to get a “normal” repository with a working tree. Use git pull inside the working directory to sync it with the server (you can set up a https://help.ubuntu.com/community/CronHowto to do this every day, for example).
?
Great!
Sorry, I was a bit imprecise: what is stored is the markdown that makes up a page’s contents (basically what one sees when one edits a page). This is definitely the most important thing. In addition there is a little bit of meta-data stored in a separate file.
The reason that it would be non-trivial to get the nLab back up from these is that one would need to re-populate the database. Things like author and revision history and so on would I think be lost, currently. Also, I’m not sure that some of the glue that makes the nLab run, like the nginx config, is backed up anywhere, at least not publically.
(To be able to get the nLab back up quickly, the best way would be to have a backup of the database, the source code in a repo, and the deployment config in a repo (possibly private). We have the second of these, but not the first or the third, though the git repository with the markdown means that at least the vital data is backed up, so that the nLab can be brought back up eventually.)
I see. Maybe the missing bits could be archived on github, too? Would be good to discuss with Adeel by email.
The entire server is also backed up to an external machine at CMU. We confirmed this with the sysadmin some time ago.
There used to be the possibility for users to pull complete local mirrors of the nLab to their local machines, and apparently there were always some people doing that, in order to have an offline version of the nLab.
If this is or could be arranged to still be possible, it would not only be useful in itself, but would also give, with enough users, an extremely robust backup system, more robust even than reliance on any institution or company could be, in the very long run.
Is the nlab history included in github backup or only the present version ?
Re: #9. Yes, I completely agree. As mentioned in another thread, I’m working on something which should make this easier.
Re: #10. Yes, from the time the backups began being made. It would need some work to extract data like the author and date of the change, but it could be done.
1 to 11 of 11