Opened at 2008-06-03T04:59:46Z
Last modified at 2016-04-03T12:14:42Z
#445 new enhancement
implement relay: allow storage servers behind NAT
Reported by: | warner | Owned by: | somebody |
---|---|---|---|
Priority: | major | Milestone: | undecided |
Component: | code-network | Version: | 1.0.0 |
Keywords: | firewall availability | Cc: | vladimir@… |
Launchpad Bug: |
Description (last modified by rvs)
Our roadmap.txt used to have "relay?" as Connection Management Step 5. This would allow storage servers to live behind NAT boxes.
Specifically, servers should be able to announce a FURL that goes through some willing relay server instead of pointing directly at the storage server. There are a couple of different Foolscap proposals to implement this (http://foolscap.lothar.com/trac/ticket/46).
Some more flexible store-and-forward routing scheme might also satisfy this goal.
Change History (9)
comment:1 Changed at 2008-06-03T05:29:21Z by warner
comment:2 Changed at 2009-09-19T22:12:29Z by zooko
If you like this ticket you may also like #754 (merge manually specified tub location with autodetected tub location).
comment:3 Changed at 2009-12-13T02:10:33Z by davidsarah
- Keywords firewall availability added
comment:4 Changed at 2009-12-13T04:03:31Z by zooko
Hm, the Upload and Erasure-Coding Helper could address this problem, if you could run the Upload and Erasure-Coding Helper on a publicly reachable host where the behind-NAT storage servers could connect to it.
I think we should close this ticket as "wontfix". I'm not longer interested in Relay. Brian, what do you think?
comment:5 Changed at 2009-12-13T04:03:51Z by zooko
- Owner set to warner
comment:6 Changed at 2009-12-13T04:04:07Z by zooko
Or perhaps closed it as "fixed", treating the Upload Helper as the fix.
comment:7 Changed at 2009-12-19T14:05:44Z by zooko
Oh wait I *am* still interested in this problem, because of this feature request: http://allmydata.org/pipermail/tahoe-dev/2009-December/003331.html . In that conversation Brian pointed out that upload-and-erasure-coding helper doesn't solve it because the current helper implements only immutable upload, and not immutable download, mutable upload, or mutable download.
comment:8 Changed at 2009-12-19T14:05:56Z by zooko
- Owner changed from warner to somebody
comment:9 Changed at 2016-04-03T12:14:42Z by rvs
- Cc vladimir@… added
- Description modified (diff)
See also #169, #49, and #50, about STUNT and hole-punching.