fix specialRemote confusion with tahoe
authorJoey Hess <joeyh@joeyh.name>
Fri, 22 Aug 2025 17:14:02 +0000 (13:14 -0400)
committerJoey Hess <joeyh@joeyh.name>
Fri, 22 Aug 2025 17:24:53 +0000 (13:24 -0400)
commit70b4220ddfe81d31cdd0976b44ac7d7d4aa293c3
tree5fa8c8a5ebdf3d2891227fcd1ce8274415f8a1a7
parent5b0cef467a933f0ef1253e59c7c811765f57dddd
fix specialRemote confusion with tahoe

tahoe: Fix bug that made initremote require an encryption= parameter,
despite git-annex encryption not being used with this special remote,
since tahoe handles encryption itself.

The chunking parameters were also accepted and won't be any longer either.
They were also not actually used.

c4ea3ca40ae6ba973287ca94e892e93973a8376e was the commit. At that point
specialRemote was being added to most remotes and I forgot tahoe doesn't
need these parameters.

Turns out that, when embedcreds=yes was used, it did *not* cause the
introducer-furl and shared-convergence-secret to be encrypted, even
though encryption= was specified. Which is only not a security hole
because encryption= was not documented to work with the tahoe special
remote at all!

It might be nice to support onlyencryptcreds=yes with tahoe, and it
would make sense to accept the encryption= parameter then, and only use
it for encrypting the creds. That would take some work, since the
encryption= parameter would need to be optional, and the usual encrypted
special remote code couldn't be used.

Sponsored-by: unqueued
CHANGELOG
Remote/Tahoe.hs
doc/special_remotes/tahoe.mdwn