PATCH samba cache and read-ahead
azez at ufomechanic.net
Fri Jan 4 13:21:33 GMT 2008
* Volker Lendecke wrote, On 04/01/08 12:55:
> On Fri, Jan 04, 2008 at 12:43:15PM +0000, Amin Azez wrote:
>> I've not looked at the implementation of the rpc over named pipes, but
>> if it works with the samba4 IDL stuff I heard about, i.e. it happens
>> "for me" and I don't have to pack and unpack arguments any more, but
>> merely write IDL function definitions them I'm definitely converted.
> Yes, it's exactly that. Your advantage is that you don't
> need to marshall anything on your own, and you don't have to
> ask anybody for name space. Just invent a named pipe name,
> run uuidgen and you're done with your local sandbox. In
> theory you could even ride on one of the existing named
> pipes, but this might be sub-obtimal.
This sounds brillig. Do you have any URLs handy with practical tips, or
any names of suitable existing named pipes for which I can examine the
related source for tips?
> Your disadvantage is that it has slightly higher processing
> costs, and that you don't have the correct tid/fid context
Err.. that sounds unfortunate.
Aren't RPC's calls made in the context of a TID?
And my calls will be proxy to proxy and within the proxy be mapped to or
from standard read/write calls anyway, so a propagatable FID should exist.
I very much appreciate you pointing this out to me, however this
suggests that to finish my current round, I should obtain a TRANS2
opcode and look at using RPCs in the future.
So, please Steve/Jeremy can I have an ID from the range? I promise not
to over-abuse it and to move to RPCs as appropriate.
More information about the samba-technical