p2p block distribution service thing. you can choose to run a server that serves some blocks (2MB chunks of data):
$ cat image.png | psync export | tee image.png.hashlist ba7816bf... cb8379ac... $ psync up localhost:8000
and then others will get data from your servers if they find you:
$ cat image.png.hashlist | pysnc get 'http://localhost:8000' $ cat image.png.hashlist | psync glue > image.png.2
there will be no mechanism for hashlist distribution, because you and
I may differ on what
image.png means. If I manage to convince myself
to a solution, maybe a checksum of hashlists then I will implement it.
at its core psync is a tool for resolving hashlists. say you have
image.png which you split into blocks and hash them: (the
checksums are shortened for brevity)
the hashes on the left are used to make a hashlist (same order as above) that you share with your friend. when your friend wants to download the file he/she:
make requests to some server for the all the chunks.
r1 = GET http://psync.io/ba7816bf r2 = GET http://psync.io/cb8379ac
for each response, check if the checksum of the contents of the response matches the expected checksum:
hash(r1) == ba7816bf (ok, keep chunk) hash(r2) != cb8379ac (something went wrong)
he/she can try to get the chunk from other servers or retry the request:
hash(r3) != cb8379ac (nope) hash(r4) != cb8379ac (nope) ... hash(rN) == cb8379ac (ok, keep chunk)
the advantage of hashlists is that retrying downloads are cheaper since you don't have to download the entire file again, and you can download files from potentially untrusted sources - the only trusted thing is the hashlist.
- come up with method for peer sharing
- gossip protocol for block discovery