Konrad, Not sure if you're aware, but SSH to a host is easy enough to parley into limited outbound Internet access. http://superuser.com/questions/370930/ssh-reverse-socks-tunnel
SSH remote port forwarding gets you SSH back out to the connecting host, and you can then use the in-built OpenSSH SOCKS proxy to make TCP connections out to the world through that machine. Native application SOCKS support is always best (eg. http://stackoverflow.com/questions/15227130/using-a-socks-proxy-with-git-for-the-http-transport), but there's also tsocks when that's not an option (http://alexborisov.org/tunnel-your-apps-via-ssh-with-tsocks/). Definitely not something you'd want to run in a script, but it would defeat the point of your firewall if you did that anyway. I guess my point is that unless you're air-gapped, you've always got connectivity options. The amazing thing about Git, IMHO, is that it's perfectly usable with an air gap too. Given the continued superiority of sneakernet for really large data (https://what-if.xkcd.com/31/) that's certainly something any good "big file" VCS would want to replicate. Cheers, Tim Tim Dettrick Senior Software Engineer The University of Queensland ITEE e-Research Group On 19/11/14 18:51, Konrad Hinsen wrote: > --On 18 novembre 2014 12:16:41 -0500 Greg Wilson > <[email protected]> wrote: > >> Thank you all for the pointers - anyone interested in doing a short blog >> post about options for the benefit of other scientists? > > Not me, sorry. I have the problem but haven't yet found any solution > yet that's worth talking about. > > For anyone who does want to tackle this, please don't forget to > mention the infrastructure requirements and limitations of potential > solutions. In my experience, the "big files" tag is usually associated > with other tags that make it hard to use version control. > > The one that bothers me most in my own work is that I use a cluster to > work on my big files, and my cluster happens to have the most extreme > firewall imaginable: no Internet access at all. I can connect to the > machine via ssh, and that's it: no other protocol, and no outgoing > connections. For me, a "big file" version control system must thus > allow repository synchronization via ssh (e.g. using scp). > > Konrad. > > > _______________________________________________ > Discuss mailing list > [email protected] > http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org > > _______________________________________________ Discuss mailing list [email protected] http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org
