Hi,
In pyobex i measure the round-trip time, and dynamically change
the mtu so that the latency is acceptable. If you set the mtu to high
it might take a _very_ long time before you can cancel the current
operation. The mtu I set is something between the minimum
mtu allowed and the mtu published by the receiver. For this same
reason I always publish a maximum mtu to the peer device, which
has to figure out itself which mtu is practical to use to (some devices
don't, like the HP capshare ;-)
-- Dag
On Tue, 6 Mar 2001 13:42:10 -0800, Jean Tourrilhes wrote:
> On Tue, Mar 06, 2001 at 09:29:33AM -0800, jt wrote:
> >
> > By the way : one of the Obex implementation there want to do
> > everything without setting a session. OpenObex deliberatly refuse to
> > do multipacket PUT or GET outside a session. Of course, the two don't
> > agree with each other.
> > Any though ?
>
> Solved this. This work perfect Linux<->Linux. This seems to be
> a bug in our WinCE Obex implementation (not setting final bits right).
>
> Another one for you :
> Obex was never geared for performance, but I believe that we
> can do better than what we are doing. Fox example, we are not even
> using the maximum packet size allowed by IrDA (~2040 bytes).
> The proposal is to define things like this :
> #define OBEX_MAXIMUM_MTU 2040
> #define OBEX_DEFAULT_MTU 1024
> #define OBEX_MINIMUM_MTU 255
> And have a call allowing the user to set the rx_mtu to any
> value he want (between MAX and MIN, initial value beeing DEFAULT). Of
> course, this call would apply only before the connection, because at
> connection time we may reduce it to what the link layer tell us.
> What do you think ?
>
> Jean
> _______________________________________________
> Linux-IrDA mailing list - [EMAIL PROTECTED]
> http://www.pasta.cs.UiT.No/mailman/listinfo/linux-irda
>
>
>
>
----
Dag Brattli <[EMAIL PROTECTED]>
My homepage http://www.brattli.net/dag/
Try Linux-IrDA: http://irda.sourceforge.net/
Try Pygmy: http://pygmy.sourceforge.net/
_______________________________________________
Linux-IrDA mailing list - [EMAIL PROTECTED]
http://www.pasta.cs.UiT.No/mailman/listinfo/linux-irda
