Re: [tlug] Peeling onions.

[ijw@example.com]

> On Thu, Mar 10, 2005 at 04:02:02AM +0900, Stephen J. Turnbull wrote:
>>     Martin> what about NFS?  filesystem dying is not exactly a new
>>     Martin> problem limited to userspace filesystems.
>> Just because the problem has a beard longer than Gandalf's doesn't
>> mean it's not worth solving.  :-)
>
> naturally,
>
> what i meant to say was either:
> shouldn't the unix world already be aware of the problem and wanting to
> solve it?
>
> or:
> if you think that this problem is not worth solving because no sane
> person would use userspace filesystems anyways then think again
>
> greetings, martin.

So:
- the problem's been around for hundreds of Unix years (*)
- NFS's solutions seem to be "keel over with some nonspecific error
message" or "hang up lots of processes".  And occasionally "refuse to
unmount" and "screw the OS over so badly that a reboot is required". 
Apparently the choice of response is implementation dependent. (**)

If yopu want Unix programs to work you have to stick with Unix FS
semantics as a basis.  I suspect that extending Unix semantics is not the
best way of soilving the problem, and a better way is to rethink your
interface entirely and then implement a compatibility layer on top, if
necessary.  Rewriting the VFS in Linux is one approach, but I'd approach
it by writing a simple test environment in which to tinker. (***)

-- 
Ian.

(*) Unix years: like dog years.  Based on how long it feels as if
something takes to run when you're sitting waiting for it.
(**) I don't claim to be an NFS expert; the above is simply my
observations of various systems in distress.
(***) "would" in this context means "won't get round to".