Re: Questions about caching on remote FS types.

[Michael Meeks <michael ximian com>]

Hi Christophe,

On Mon, 2002-10-28 at 10:29, Christophe Fergeau wrote:
> Is this daemon doing anything useful atm ?

	Nope ...

>  or is it just a proof of
> concept which just starts when necessary and dies when necessary ?

	Exactly that.

> Would this daemon take care of displaying dialogs asking for passwords,
> or would it be necessary, or should it be done elsewhere ?

	No - that would be done by a client, assuming there was no password
stored in there; of course - useful as the daemon is for file caching
the password stuff is rather more contentious security wise - inasmuch
that it'd be nice to restrict which apps can access the information, and
that requires some ORB level grunt really.

>  I know there
> is noone 'officially' working on that daemon, but I'd like to have an
> idea on what is your vision on what it should/shouldn't to.

	My vision is that we move slow protocol backends into the daemon, eg.
'ftp', 'smb' etc. and de-couple them from the client via. a nice CORBA
interface that mirrors / wraps the existing file / directory API [ this
in itself is rather a painful amount of boilerplate wrapper typing ].

	Then at least without being able to ask for passwords back, we can
store them centrally [ much less controversial ], and better - cache /
timeout lots of resources, such as ftp connections, smb: browse
information, downloaded files - and crucially [ why I was originally
interested in this ], deal with archive files efficiently, by
maintaining a decent lifecycle for the necessary caching bits.

	Is there any of that you'd like to look into ? If you do the tedious
IDL interface wrapping stuff for VFS API <-> IDL types then I'm happy to
continue my poking to make the ORB client code thread-safe and it easier
to write async. server impl's. [ for cancellation, debuggability etc. ].

	How does that sound ?

	Regards,

		Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot