[tahoe-dev] up with filesystems! up with the web!

Chimpy McSimian IV, Esq. mr.monkey at gmail.com
Tue Jan 12 14:17:12 PST 2010


Although Octavia's Java implementation arguably nearly does semi-work
somewhat as an ugly test harness, nobody actually uses it right now.
However, when/if it is a real filesystem (Probably via Fuse), it will
be notable for completely ignoring POSIX and supporting only a small
subset of normal filesystem features. There are no symlinks, nor even
multiple hard links; there is only mtime (be honest: do you really
care?); there are no ACLs or permissions.

Software philosophers and people who get lots of mileage out of NTFS
alternate streams, resource forks, and WebDAV extensible metadata
schemata are going to hate it. I'm ok with that. :)



On Sun, Jan 3, 2010 at 11:57 PM, Zooko O'Whielacronx <zookog at gmail.com> wrote:
> This has been an interesting conversation to read.  Thanks, Chimpy,
> David-Sarah, and James.
>
> It sounds like this is one of the "deep" ways that the design of
> Octavia differs from Tahoe-LAFS.  I would be interested in how people
> use Octavia in practice.  (I understand that it is currently at a very
> early stage of development.)  I'm also interested in other ways that
> the design of Octavia differs from that of Tahoe-LAFS.
>
> I must confess I don't understand how it would be possible to
> implement James's suggestion for a tree-shaped directory structure.  I
> suppose it would require that your client be on the lookout for cycles
> or converging links that had been created by a buggy or malicious
> client or by some sort of accident caused by partition, and raise an
> error if one is detected.  I suppose this would be analogous to the
> way that fsck looks for such problems in local filesystems.
>
> Hm, I also suppose that we'll need to implement something like this
> sort of detection anyway in order to facilitate deep-copying between
> Tahoe-LAFS which has arbitrary graph structure and local filesystems
> which have (various) limitations on the structure: #104, #641, #850.
>
> Regards,
>
> Zooko
>
> http://allmydata.org/trac/tahoe/ticket/104# does cp -r work as expected?
> http://allmydata.org/trac/tahoe/ticket/641# tahoe backup should be
> able to backup symlinks
> http://allmydata.org/trac/tahoe/ticket/850# tahoe backup loops on
> recursive links
> _______________________________________________
> tahoe-dev mailing list
> tahoe-dev at allmydata.org
> http://allmydata.org/cgi-bin/mailman/listinfo/tahoe-dev
>


More information about the tahoe-dev mailing list