[gridrpc-wg] Final call: GridRPC API for inclusion in SAGA 1.0

[Andre Merzky]

Yusuke, 

Quoting [Yusuke Tanimura] (Aug 10 2006):
> 
> Hi,
> 
> > >   - The RPC spec is silent about 'when' the connection to
> > >     the remote server is made, on creation of the handle, or
> > >     on call().  We decided in other parts of the spec that,
> > >     for example, the constructor opens a file, or remote
> > >     connection.  I propose to prescribe the same for RPC.
> > >     Does that make sense?  Do we need to loosen the
> > >     semantics elsewhere in the spec? (IMHO not).
> > 
> > I am afraid, doing so would impose restrictions on the underlying gridrpc
> > implementations. some of them might not meet them, but would still be useful
> > if SAGA could access them.
> > 
> > What you really would have in mind is a test "can execute" of the procedure?
> > I am not sure if the constructor could aalready give a comprehensive answer???
> > 
> > Hidemoto, Yusuke, any ideas ???
> 
> In case of the GridRPC, when the connection to the remote server is
> made is free for each implementation.  In the constructor, the server
> will be certainly assigned for the RPC.  The connection for data transfer
> may be established in the constructor (Ninf-G) or may be done in
> rpc.call (GridSolve).

Right, that is what I remember from last GGF as well.  That
means effectively that the call() method must be able to
throw all exceptions we have on the constructor of the
handle.

No, the question is how do we handle that for other SAGA
packages, e.g. for files?  Here the constructor implies an
open(), and could return DoesNotExist.  If SAGA would allow
to delay the opening of the file, it would be possible to
save a round trip time, and to optimize the implementation -
but the drawback is that on _any_ other call (read(),
write() etc) we could meet a DoesNotExist exception.

I guess we don't want that, and should limit that semantics
to RPC.  However, we then end up with an inconsistency in
the spec (file constructor behaves differently than rpc
constructor), which we need to document explictely (at
least).

Does that make sense?

-- 
"So much time, so little to do..."  -- Garfield