tim@hoptoad.UUCP (Tim Maroney) (11/16/86)
We are currently shifting from Appletalk protocols to Internet protocols with our remote function protocol, which now runs on top of ATP (gag). We want to be able to port quickly to various operating systems. My question is whether RDP is quickly catching on. It's currently listed as a minor host protocol; is it starting to become a major host protocol? I would like to run RFP on top of RDP, which seems perfectly suited to the task, but it is likely we will use TCP instead if RDP is still fairly obscure. What about UDP? Well, I don't feel like adding yet another single-protocol reliability layer to it! I think RDP is what UDP should have been in the first place.
braden@ISI.EDU (Bob Braden) (11/19/86)
I believe one can say that RDP is not widely implemented or used. In my opinion, it should be regarded as an experimental protocol. There are two good ideas in RDP (selective retransmission and packet-orientation), which are being incorporated into more recent experimental protcols -- eg VMTP and NETBLT. Note that packet orientation is also a feature we will all obtain eventually from TP4. I have also heard some opinions (from within BBN as well as elsewhere) that there are some not-so-good ideas in RDP. In my opinion (again), if RDP were to become seriously used, it would need a cycle of improvement. It would not seem to be a big win over TCP in most applications. Bob Braden
walsh@HARVARD.HARVARD.EDU (Bob Walsh) (11/19/86)
Craig, I understand that you are at least referring to my code and have looked at in writing your "independent" implementation. bob
craig@LOKI.BBN.COM (Craig Partridge) (11/20/86)
Tim, I know of only two implementations, neither of which is in general circulation. I'd be interested in hearing of others. The first was done by Bob Walsh while he was at BBN, and is part of the BBN TCP/IP implementation. The second is the one I'm currently working on (despite my BBN affliation I'm doing this independently of Walsh's implementation, as work towards a M.Sc. degree at Harvard). Right now I'm worrying about benchmarking and evaluating the implementation and, as yet, have given no real thought to questions of distributing it. Both implementations are for the 4.2/4.3bsd distributions. Craig
craig@LOKI.BBN.COM (Craig Partridge) (11/20/86)
> I understand that you are at least referring to my code and have looked > at in writing your "independent" implementation. Bob, I've used your implementation for conformance testing (comparing results of checksum routines, opening connections between implementations, etc). So yes, I've leveraged off your work a bit. However, for a variety of reasons, I did write my implementation from scratch -- not one line was taken from yours. I think that qualifies as independent, without the quotation marks. It seemed to me that using another implementaton for conformance testing was only reasonable practice and did not require explicit acknowledgement. If you feel somehow slighted, my apologies, that was certainly not my intent. Craig