Arkady
Wed Jan 19 05:11:07 CST 2005
Thanks, Ali !
That really was I wrote about : the NIC halted after reset to half-duplex
and restarted. Then program ( the same ) send the data ( the same ) with
TCPWindowSize set to 17K ( start size ) instead of 64K ( start size too )
in full-duplex.
Arkady
"Alireza Dabagh [MS]" <alid@online.microsoft.com> wrote in message
news:Ocfzhzb$EHA.3416@TK2MSFTNGP09.phx.gbl...
> TCP/IP has no way of knowing that duplex has changed (not today). If you
> change any advanced property of a NIC, you will cause the NIC to get
halted
> and restarted again in which case all TCP connections will be lost. The
> change that you in TCP window size might be more related to reestablishing
> the connection than anything else.
> Even if you could change the effective link speed of the network without
> restarting the NIC, (unplug from a 100meg switch, plug to a 10 Meg switch)
> and you noticed a change in TCP window size, you can not assume that TCP
> window size change is due to stack dynamically (=after the bind process)
> "directly" monitoring the reported link speed from the NIC. Any "dynamic"
> changes here -could- be the result of the self regulating algorithm in the
> stack more than anything else.
>
> NDIS 6.0 will support asymmetric link speed.By then, hopefully how link
> speeds will be used for what purpose (routing table metrics, TCP initial
> window size/connection time out determination, etc.) will be documented
> somewhere.
>
> -ali
>
> --
> This posting is provided "AS IS" with no warranties, and confers no
rights.
>
> "Arkady Frenkel" <arkadyf@hotmailxdotxcom> wrote in message
> news:exXT%23U0%23EHA.2596@tk2msftngp13.phx.gbl...
> > You can check that in any modern Os ( including XP SP2 ) : set
half-duplex
> > and you'll see the change in TCPWindowSize by any sniffer ( netmon ,
> > ethereal... ). Unless Intel still continue to produce serialized drivers
> > :)
> > Arkady
> >
> > "Maxim S. Shatskih" <maxim@storagecraft.com> wrote in message
> > news:#FujTev#EHA.3472@TK2MSFTNGP14.phx.gbl...
> >> Alireza already told us that the "full-duplex" mode is a no-op in
> > modern
> >> Windows, since the real deserialized miniports are out here.
> >>
> >> --
> >> Maxim Shatskih, Windows DDK MVP
> >> StorageCraft Corporation
> >> maxim@storagecraft.com
> >>
http://www.storagecraft.com
> >>
> >> "Pavel A." <pavel_a@NOwritemeNO.com> wrote in message
> >> news:%23CSP5Xq%23EHA.600@TK2MSFTNGP09.phx.gbl...
> >> > ?? Where that duplex mode is set in the registry?
> >> > All device specific options like duplex, media type etc are not
> >> > relevant
> > for
> >> me.
> >> >
> >> >
> >> > "Arkady Frenkel" <arkadyf@hotmailxdotxcom> wrote in message
> >> news:umET6gj#EHA.2584@TK2MSFTNGP09.phx.gbl...
> >> > > Forgot to mention. That you have to take radio/line speed not
current
> >> > > up/down.
> >> > > There are a lot for 802.11 for sure
> >> > > Arkady
> >> > >
> >> > > "Arkady Frenkel" <arkadyf@hotmailxdotxcom> wrote in message
> >> > > news:OsyISBj#EHA.3472@TK2MSFTNGP14.phx.gbl...
> >> > > > Pavel !
> >> > > > Be aware that speed changed due to duplex mode set from/in
registry
> > too
> >> > > > Arkady
> >> > > >
> >> > > > "Pavel A." <pavel_a@NOwritemeNO.com> wrote in message
> >> > > > news:eq1tQre#EHA.3708@TK2MSFTNGP14.phx.gbl...
> >> > > > > Dear NDIS gurus,
> >> > > > > Now I have adapter with asymmetric send/receive speed ( uplink
> > speed
> >> > > about
> >> > > > 1/3... 1/4 of downlink).
> >> > > > > The media type is ethernet (generic or, possibly, 802.11
> > emulated).
> >> > > > > How TCPIP relies on link speed value reported by driver? which
> > link
> >> > > speed
> >> > > > the driver should report?
> >> > > > > And same question, if the device does throttling and
> > prioritizing -
> >> the
> >> > > > actual link speed
> >> > > > > is different for different payload types.
> >> > > > >
> >> > > > > The marketing, sure enough, wants to display the biggest speed
> > value...
> >> > > > >
> >> > > > > Thanks,
> >> > > > > Pavel
> >> > > > >
> >> > > > >
> >> > > > >
> >> > > > >
> >> > > >
> >> > > >
> >> > >
> >> > >
> >> >
> >> >
> >>
> >>
> >
> >
>
>