Re: [Usability] Medusa questions (was: Re: New way of accessing software)

["Curtis C. Hovey" <sinzui cox net>]

On Sat, 2003-06-14 at 19:56, Daniel Borgmann wrote:
> Does Medusa update the "index" (if that's the name) in realtime or will
> it do incremental updates like updatedb? I assume that the former might
> be possible with FAM.

Getting incremental indexing working is a high priority.  It uses cron
to recreate the entire index right now.  Using FAM seams a given, but it
has varying performance on different architectures which may require so
indexing strategies to overcome.

> How does Medusa compare to "native" metadata supported by the filesystem
> (like BeFS) or a database filesystem? Especially in regard to
> performance?

Medusa is unaware of native file system meta data as it uses gnome-vfs
to normalize all access.  Adding new features to gnome-vfs would be
key.  Medusa uses DB1 to store it's data.  I believe that once Medusa is
re-integrated into Gnome we should change the back end.  Seth and
Rodrigo have suggested SQLLite/libgda--an excellent suggestion.  I'd
like to make changes to the scheme at this time to add ranking
information to provide sorted results that best match the search
criteria.  This is also the time to look at storing additional file
system meta data.

> Once filesystems get this functionality (Reiser4?), would it
> theoretically be possible for Medusa to transparently make use of those
> features if available? And would it make a difference anyway?

It would make a difference for the better.

> And last but not least, do you have any opinion about Microsoft's
> "Future Storage" plans? :) Do you know enough about this to somehow tell
> how this compares to Medusa and/or filesystems like Reiser4?

Architecturally, they are all very different from each other.  They all
share the same goal of making files more manageable to users and
applications, but they are solving different problems.  

Medusa is a user service that is compatible with a very diverse set of
OSes and file systems.  I'm sure a system level service will be created
when Medusa gains momentum.  Being an addon service, there is extra cost
keeping the index current, but it is also easier to extend when needed. 
As all file systems are treated equally, it is ideally suited to
enterprise and user network computing.

Reiser is a system level service that run on a smaller range of OSes. It
has excellent performance, and application can can some benefit from it
without knowing it exists.  But in enterprise and user computing, a
machine is connected to a diverse set of file systems, (SMB, NFS, NTFS,
UFS, etc.) and only a small subset of files will have the additional
meta data that users may depend on.  In this scenario, some files simply
cannot be have meta data.

Longhorn's SQLServer DB is quite a grand plan.  It only runs on one
architecture, and it is too large and complex to replace the file
system--there is will still be one to bootstrap the system.  I believe
that the DB will save meta data about connected file systems to some
degree, and will be flexible enough to create ad hoc attributes to
files.

-- 
__C U R T I S  C.  H O V E Y____________________
sinzui cox net
Guilty of stealing everything I am.