Steve Eirium <[hidden email]>
DISCUSSION
RAID5 is well-understood but under-utilised, mostly due to the massive logistical problems of doing anything with it. RAID is a layer below the filesystem and above the disk partitions, one would have to be some kind of pervert to bother with such arcane nonsense. Just google around to see what a wasteland of human experience it is.
The reality of the security you get on the life of your data is very simple and there's no need to have it hooked up as arbitrary chunking of your disk under the filesystem. It would be much less expensive and more classy to use files themselves as our chunks.
So what I'm proposing to build is a multi-master database for tracking the member filesystems. 1/(n-1) of each filesystem would be the node's share of the parity.
Anyway, your laptop gets stolen but the two disks in your PC have it all between them, so even though you've got nearly 400G of shite and 600G of disk, you're OK.
DELIVERABLES
Something for scanning a filesystem into a git database, on cron or whatever, depending on mtime and/or hashing.
Some more of that same thing for trying to talk to the other instances. This could be as flexible as git, I suppose.
Something else for talking to users, getting set up, recovering lost filesystems.
A project page with a funky domain name so it can really thrive out there.
PROJECT DETAILS
...
PROJECT SCHEDULE
2 months of slightly involved hacking.
BIO
http://github.com/st3vil was writing perl professionally for three years, but now only wants to do it for good cause. I have another robotical project and I think they will share some intellect, if not actual code.
I'm an unemployed musician, mostly by mental necessity. I feel this is my job as a human, I just need a way to get off social welfare (I have one in the pipes). It would be great to get some bucks for staying alive. I live in a garage fairly cheaply right now but I'll probably live in my car again in a few weeks what with it being summer.
AMOUNT
$2000 please.
Sounds like a reasonable idea. Although it doesn't directly improve Perl, it does make a useful feature available through Perl.
How will this be different from mogilefs, which is already primarily in perl (and has a substantial community behind it ...?)