>> From dshane@qual-pro.com Mon Dec 14 20:26:38 1998
>>
>> In an adhoc network, suppose that the hardware required to deliver a single packet,
>> which has no relationship to past or future packets, (ex., UDP) is less than
>> the hardware needed to support a connection based protocol (such as TCP).
>> Would that influence your position; if so, how?
I am afraid I do not quite understand the question. If the issue is UDP versus TCP,
I am not sure if the needed hardware would be different. What sort of hardware
capability differences you had in mind? (I can only think of possible differences
in source coding.)
- nitin
>> Thanks,
>> Darrell Shane
>> dshane@qual-pro.com
>>
>> -----Original Message-----
>> From: Nitin H Vaidya [SMTP:vaidya@cs.tamu.edu]
>> Sent: Monday, December 14, 1998 5:00 PM
>> To: Emad.Qaddoura.emadq@nt.com; pilc@lerc.nasa.gov
>> Subject: RE: TCP over wireless type links
>>
>>
>> >> From Emad.Qaddoura.emadq@nt.com Mon Dec 14 17:32:19 1998
>> >>
>> >> My point was, there are so many link layers. These link layers run
>> >> other
>> >> transport than TCP and TCP would be one of the options. Do ALL of
>> >> these
>> >> link layers adapt/optimize their setting for TCP or would it be
>> >> easier if TCP
>> >> does the adaptation?
>> >>
>> >> Yes, I agree, optimizing both would be the best of all, but
>> >> most likely would
>> >> not really happen, just my opinion.
>> >>
>>
>>
>> I agree with you on this. I do not think making TCP-specific
>> changes to the link layer is the best choice (i.e., the link
>> layer should not have to know that the packets it is
>> transmitting are TCP packets).
>>
>> Adapting link layer for TCP does not necessarily mean TCP-specific
>> adaptations. "TCP-unaware" adaptations would be more useful
>> (these would be unlike the TCP-aware link layer protocols that
>> have been proposed previously).
>>
>> TCP-unaware schemes can
>> benefit TCP (and potentially other traffic too). To give
>> a trivial example, using FEC on a wireless link to reduce
>> the error rate may benefit TCP, without the link layer
>> being TCP-specific (or TCP-aware).
>>
>> By the same token, if we develop some interface that specifies
>> to the link layer that a given packet should be retransmitted
>> at most X times, then the X can be chosen appropriately by the
>> upper layers (for instance, for audio, X may be 0,
>> for TCP, some non-zero number, and so on).
>> Again, to do this, the link layer does not have to TCP-aware.
>>
>> Regards.
>>
>> - nitin
>>
>>
>>
>>
>> >>
>> >>
>> >>
>> >> -----Original Message-----
>> >> > From: Nitin H Vaidya [SMTP:vaidya@cs.tamu.edu]
>> >> > Sent: Monday, December 14, 1998 1:38 PM
>> >> > To: Qaddoura, Emad [RICH2:IP10-M:EXCH]; pilc@lerc.nasa.gov
>> >> > Subject: Re: TCP over wireless type links
>> >> >
>> >> >
>> >> > >> From Emad.Qaddoura.emadq@nt.com Mon Dec 14 12:59:42 1998
>> >> > ...
>> >> > >> It would also be hard to imagine that many link layers will adapt to
>> >> > >> TCP requirements. It would
>> >> > >> seem more of a managable problem to have TCP adapt to different
>> >> > link
>> >> > >> behaviors. Therefore, the
>> >> >
>> >> > Can you elaborate why it is more manageable to change TCP as compared
>> >> > to the link layer. What if link layer adaptation is sufficient
>> >> > to yield decent performance - should we still adapt TCP instead?
>> >> > I like to think not.
>> >> > (granted that link layer alone may not yield optimal
>> >> > performance in all scenarios, but it would in some).
>> >> >
>> >> > - nitin
>> >>
>>
>>
>>
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 09:12:19 EST