DVFS and Metadata
elanthis at awesomeplay.com
Mon Mar 7 21:38:36 EET 2005
On Sun, 2005-03-06 at 16:13 +0100, Timo Stuelten wrote:
> Does it make sense to integrate some Metadata-Query-Thing (MQT) like
> Spotlight/OS-X into the VFS-Layer? If reading or creating new files, that
> MQT could then read contents, structures, metadata, whatever on-the-fly
> without the need to create copies.
I do believe that a standard query language would be very nice, and
absolutely essential once we have file systems based on meta-data
queries. However, at this point in time, I don't think it's worth
focusing overly much on. First, we don't have a stable file system that
can really make use of these capabilities. The systems that are around
today are very experimental, incomplete, etc.
I think such developments should occur outside D-VFS, and once meta-data
based file systems are in a usable state, extending the D-VFS to provide
rich query capabilities should then be done.
Adding a query language now that is based mostly on speculation or on a
single experimental system's semantics will not be likely to provide a
generally usable API or query language.
In short, while it will be some time before D-VFS 1.0 is available, I
don't think we should target meta-data query as a prime feature just
yet. If something is usable by 1.0, great, but we should accept punting
the query language to after 1.0.
It's also possible to perhaps put a meta-data engine over D-VFS. For
example, in a system based entirely on querying file meta-data, I
wouldn't expect to have to worry about browsing to a specific share. A
meta-data system could use D-VFS to pull file contents both from the
local disk and from remote shares to build the cache/index, and wrap the
D-VFS API in such a way as to make the actual shares and paths
completely hidden to the developer and user.
Sean Middleditch <elanthis at awesomeplay.com>
More information about the xdg