[tahoe-dev] FUSE, WebDAV, SFTP Re: barriers to using tahoe

[Zooko Wilcox-O'Hearn]

Dear Ed Pimentl:

On Thursday, 2010-02-04, at 20:43 , EdPimentl wrote:

> Hello Distinquished members of this list.
>
> I have decided to double ($800.00 USD) the bounty for:
>  - FUSE like layer
>  - SMB by default for Windows
>  - SFTP by default (localhost) for *nix

That's really great that you're offering bounties for Tahoe-LAFS  
improvements. Thank you!

But do you mean that you'll pay someone $800.00 USD when they've done  
all three of those things, or any one of them, or what?

By a "FUSE like layer", what do you mean exactly? I guess you mean  
that you can navigate and manipulate your Tahoe-LAFS files using your  
standard tools: fopen(), fread(), etc. in C and ls, mv etc. in a  
shell. Is that right? Let's write down exactly what the criteria are  
in a ticket so we can tell when we've succeeded. Note that we already  
have tests thanks to Nathan Wilcox, Rob Kinninmont, François  
Deppierraz and David-Sarah Hopwood which can tell us how complete a  
FUSE implementation is:

http://allmydata.org/trac/tahoe/browser/contrib/fuse/runtests.py

So, other ways to achieve the same thing -- fopen(), fread(), ls, mv  
-- would be to implement SFTP and use sshfs (http:// 
fuse.sourceforge.net/sshfs.html ) to provide the filesystem  
interface, or implement WebDAV and use davfs2 (http:// 
savannah.nongnu.org/projects/davfs2 ) to provide the filesystem  
interface.

Now it turns out that we already have an SFTP implementation, but it  
lacks test and some people claim that it doesn't work for them:

http://allmydata.org/trac/tahoe/ticket/531

Alberto Berti contributed a patch to make it work for him, but then  
Ryan Heimbuchs said that this patch made it break for him.

So the next thing we need is tests of Alberto's and Ryan's issues.

Regards,

Zooko