[tahoe-dev] barriers to using tahoe

[Jody Harris]

On Thu, Feb 4, 2010 at 4:34 PM, Nathan Eisenberg <nathan at atlasnetworks.us>wrote:

> > Users expect a / directory in a filesystem. Tahoe isn't a filesystem, but
> it does things a filesystem does (store and
> > retrieve files). They expect to interact with it like a filesystem
> because they see it as a filesystem.
>
> That would seem to conflict with the expectation set by the name 'Least
> Authority File System'.
>
> Understand, I'm not being catty at all - I'm just saying that if this
> expectation conflict is confusing to me (and I've spent weeks reading wikis
> and mailing list history and experimenting with test grids), it will
> probably be confusing for others as well.
>
> Best Regards,
> Nathan Eisenberg
>

Nathan, I'm beginning to see the problem.

Basically it boils down to this:

The "average user" will attempt to interact with Tahoe-LAFS in a way that
violates the design parameters of Tahoe-LAFS.

The only way to "fix" this "bug" will be to break the "LAFS" part of
Tahoe-LAFS.

Perhaps a "independent subsidiary" project would break LAFS by creating a
metaphor allowing users to interact with Tahoe as a filesystem.

I do not have the understanding to undertake such a project, and I shudder
at the prospects of gaining that understanding....

I remember being surprised when I first realized that a child directory did
not have a '.' link back to it's parent. That was when I began to realize
that Tahoe-LAFS is not a filesystem. There was a "hey, this is a bug
moment," followed by an, "oh, yeah," moment.

I can't wait around for Octavia to materialize, and I don't want my Tahoe
grid to fracture by defection. I'm stuck between two worlds....
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://allmydata.org/pipermail/tahoe-dev/attachments/20100204/22ee4515/attachment.htm