>> I don't think its just a few people who think that way. Unified,
>> globally scalable services using a distributed file system and
>> single-login authentication is the model which Novell,
Novell does try to do everything through the filesystem. But globally
scalable? I think not.
>> Microsoft,
Let's see what Microsoft has done with networking. Windows for
Workgroups doesn't have a central distributed filesystem. And its
other services aren't based on a filesystem model. NT is only
intended for servers, so it's got to use other protocols to do
anything real.
>> the OSF
OSF provides a distributed filesystem, but the framework is based on
DCE RPC and the location services, not on the filesystem. This is a
tractable model, but it isn't the one you're pushing.
>> and probably most other significant computer network players have
>> adopted as a strategic architectural direction.
Oh, you mean like the IETF, the folks who bring you the internet
protocol suite? Even OSI, home of the Bloated Protocol, doesn't try
to do everything through the filesystem.
>> In a nutshell, distributed services are scalable and manageable -- two
>> critical features that are missing on most large networks today.
Actually, most distributed services I've seen aren't as scalable or
manageable as I would like. But using the filesystem as the
fundamental infrastructure isn't going to give you scalability or
manageability. It's going to give you a mess.
>> Maybe its silly to put /usr/spool/mail within AFS. But the concept of
>> distributing a campus mail service is rich and worth pursuing.
I agree. But doing it through the filesystem is bogus.
Marc
