I'd be interested to know what semantics are lost
by intermediate acknowledgements. When I look at the
WinSock socket interface, I don't get any notice after I do a send
of whether the data I sent actually arrived. This has to happen
(I think) at the application layer.
    Doug.............
Ketan Bajaj wrote:
> Hi Arkesh,
>
> I  may be able to answer your first question. Yes, a buffering policy can
> be used at hub, but the hub CAN NOT send back acks to the Tcp data source
> as it would violate the TCP semantics. This is because data has not
> actually reached the correct destination.
>
> I have been working on a link layer protocol for improving performance of
> TCP over satellite links, which uses buffering concepts and a number of
> state variables for maintaining semantics and handling other issues.
>
> The report and the performance measurements will be ready by December
> 15th. I could send a copy then ...
>
> Regards,
>
> Ketan Bajaj
>
> Department of Computer Sci.& Engg.
> Indian Institute of Technology
> New Delhi
>
> On Fri, 30 Oct 1998, R.S.Arkesh Kumar wrote:
>
> > Hi
> >       We have started work on a project "Internet Access via Satellite".
> > The basic architecture is that of a TDM/TDMA, with the interface to the
> > Internet at the Hub being a standard Router and a Internt Server(a
> > Workstation).
> >       I have the following doubts regarding TCP/IP implementation over
> > the Satellite :
> > 1. Why is TCP/IP implemented over the Satellite ? Cannot something like a
> > "Store and Forward" policy be adopted, i.e., all the Internet datagrams
> > which land at the Hub are buffered by the server, and it itself generates
> > the ACK packet to the source.  The server will then encapsulate the data
> > in a proprietary format and sends it to the appropriate VSAT(remote
> > terminal).  Any data lost in the satellite link will be taken care by the
> > server at the Hub by retransmission independent of the original source.
> > The same policy is used for the IP datagrams generated by the VSAT.
> >       Wouldn't the above method be a better implementation, rather than
> > running TCP/IP over the satellite ?
> >       Can the same TCP/IP Protocol stack as used for terrestrial
> > networks be used in this case ?
> >
> > b. Is a back channel of 64 kbps which will be shared by a number of VSATs
> > on a TDMA be sufficient for the ACK packets generated by TCP ? Will a
> > smaller data rate affect the overall throughput ? Where will I get more
> > details in this regard ?
> >
> > c. Can a Router be configured for a network MTU manually ? / How does a
> > Router learn the MTU of a connected network ?
> >
> > Reply ASAP,
> > With regards,
> > Arkesh
> > ###############################################
> >
> > R. S. ARKESH KUMAR
> > RESEARCH ENGINEER
> > SATELLITE COMMUNICATION
> > CENTRE FOR DEVELOPMENT OF TELEMATICS
> > 71/1, MILLER ROAD
> > BANGALORE - 560 052
> >
> > Other E-Mail address : [email protected]
> >
> > PHONE : 91-80-2261529
> > FAX:91-80-2263256
> >
> > ################################################
> >
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:48 EST