The RFB server is started upon image start (accepting only local connections). I like your idea, but it starts to become a cumbersome process if one must 1) forward the port, then 2) send a kill signal, then 3) open their VNC viewer to do the work, THEN 4) remember to do some other step to ensure they stop the RFB server.. Probably, the last step would be forgotten and we're right back to it running 24x7..
What I tried to suggest was to keep it running. I don't use the signal trick, because I don't see any problem with letting RFB run.
Ah, good.
Our existing repository is currently 17GB worth of mcz's and the Magma copy may need about the same by the bulk-load is done..
We can make a few more GBs. Btw, I thought Magma wouldn't copy the mczs.
It doesn't copy mcz's, but it does open them up to get the MCVersion object inside, and adds it (along with all of its referenced MCDefinition objects) directly to a big Collection in the database. The MCDefinitions are canonicalized, so there is some space savings there, however, the source code is not compressed like in an MCZ, so the amount of space it will ultimately require is very hard to predict.
I'm keeping my eye on it, and since we don't need any of the files in the /home/squeaksource/webserver/ss.magma/commits directory, a lot of space is recoverable from there, so I think we'll be fine..
Best.