[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#833090: ITP: magic-wormhole -- Securely and simply transfer data between computers



On Mon, Aug 8, 2016 at 4:42 PM, Antoine Beaupré <anarcat@debian.org> wrote:
It is not hardcoded: you can change the central host and relay with a
simple commandline option.

$ wormhole --help | grep -B1 'to use'
Options:
  --relay-url URL                 rendezvous relay to use
  --transit-helper tcp:HOST:PORT  transit relay to use

This could, arguably, be done in a configuration file to facilitate
using third party servers, but this can hardly be considered
hardcoded. Anyways, if the current main host goes down, I assume the
software can/will be patched to provide other hosts as options.

Keep in mind transfers are ephemeral: the central hosts are used only to
establish contact and transfer the file, then everything is torn down.

Fair enough, *a* central host is hardcoded. You could obviously set up your own, which sort of defeats the purpose of being simple, but point conceded.
 
>> We still ship FTP daemons that serve files without passwords and use
>> cleartext by default.
>
> They're not labeled "secure" though ;)

Actually, quite a few are:

[ ... snip ... ]


I agree it is somewhat of an empty word, but it shouldn't be considered
reason enough to keep stuff from entering Debian, because then you'd
have a *lot* of packages to kick out the archive. Heck, "apt search
secure" suggests I installed zendframework, and we know how scary PHP
security has been in the past. ;)

Good point.
 
> Just to clarify, I never objected to the package itself, just that I
> wasn't sure about it being called "secure".  I don't know enough about
> the algorithms and attack surfaces involved to make any kind of
> qualified statement though, so maybe it does qualify as secure.

Well, I am not a cryptographer myself, so I can't comment about the
algorithm. But I am somewhat familiar with such protocols and I found
they brought a novel and robust system in place, that has similar
robustness properties than existing protocols (e.g. Oauth with a
digit-only PIN) with interesting enhancements that make it fail more
gracefully (abort transfer after first failed attempt).

May I suggest that, if you do not know enough about security protocols,
you refrain from discouraging people, that do have some knowledge about
them, from packaging software into Debian? :)

Will do. Crawling back under my rock.

At no point was I attempting to discourage anyone, apologies if it came off that way. It does look like a really useful tool, I just figured it might need a tiny bit more vetting before calling it secure. It appears I was wrong, which is cool. I like being wrong. :)

Cheers,
Fredrik.

Reply to: