lauren@vortex.UUCP (Lauren Weinstein) (08/18/83)
I certainly have nothing against CSNET, nor do I have any direct connection with the organization at this time. Any impression I've gotten of "problems" with CSNET services is from netmail and phone calls I've gotten from people who have related various problems to me over a period of time. This is not to obscure the fact that many sites have found CSNET to be useful for the "mail-only" purpose that it serves, though there are clearly some problems. By the way, my personal feeling is that Usenet potentially offers as reliable a gateway to ARPANET as CSNET, once Usenet users are properly educated. There is just no way that the gateways will ever be completely shut down in any normal set of circumstances, since too many ARPA people find it extremely valuable to communicate with persons at Usenet sites. If some sort of serious ARPA "breach" ever occurred, then we might see a gateway shutdown, but the odds are that ALL external nets would get knocked off, probably including CSNET as well in such a case. This is a personal opinion. Very briefly, the main complaints I hear about CSNET are: 1) Cost. Both the fee to join and the transmission costs (remember that sites generally talk only to official "relay" nodes -- the net is not distributed in the "localized" manner of Usenet) are of concern to a considerable number of persons, apparently. 2) Dealing with the Organization. Apparently the the typical problems of dealing with a "central" controlling organization are present in many situations with CSNET. 3) Protocol problems. The dialup communications protocol used is extremely crude (for example, it uses an actual checksum, not even a real CRC like the UUCP packet protocol) and does not support reasonable "windowing" (again, unlike UUCP). I have been informed of easily triggered fatal "deadlock" conditions in their dialup protocol, though I haven't had a chance to personally verify this last point yet, but I plan to try soon. 4) Bug fixes. CSNET tries to maintain tight control over the modification of sources relating to the net. This may well lead to more standardization, but MANY users report that bug reports are not acted upon for weeks or months or longer. The CSNET/MMDF code is quite voluminous and rather complex, so this can become a serious problem. I've heard *very* many complaints on this issue. I will freely admit that most of the above is not based on personal experience (except for some issues regarding the protocol, with which I am directly familiar) and is instead based on reports given to me by others. Feel free to accept or reject this information as you see fit. --Lauren--
mark@umcp-cs.UUCP (08/19/83)
1) Cost. One reason Csnet costs, which is also the reason why its arpa connections are so much more reliable than usenets, is that Csnet pays approximately $1M/year to Darpa for the privilege of being on the arpanet. This cost is paid for by NSF. The other reason is that every site is one hop from every other. This mean message get transmitted with fixed delay. 2) Dealing with the Organization. I have not found this to every be a problem. There is a 24-hour hotline number, as well as very good responsiveness to mail. When something goes wrong and you are the only hacker around and you are just not sure exactly what to do but you suspect mmdf (the Csnet software) needs attention, it is nice to have someone to talk to. 3) Protocol problems. True. 4) Bug fixes. CSNET tries to maintain tight control over the modification of sources relating to the net. I've heard *very* many complaints on this issue. This I really do not understand. If Csnet really wanted to maintain tight control over the sources they would not give them out. We have modified ours extensively, and the Csnet crowd reaction ranges from who cares to that's interesting, send us the code. Neither is tight control, whatever that means. I will freely admit that most of the above is not based on personal experience (except for some issues regarding the protocol, with which I am directly familiar) and is instead based on reports given to me by others. Furthermore, it is only about the protocol that you knew what you were talking about. I think your informants had an ax to grind.-- spoken: mark weiser UUCP: {seismo,allegra,brl-bmd}!umcp-cs!mark CSNet: mark@umcp-cs ARPA: mark.umcp-cs@UDel-Relay
whm@arizona.UUCP (08/24/83)
If CSNET (via NSF) is paying DARPA $1M per year for the connection to ARPANET, wouldn't that seem to chill any hopes of legitimitizing a Usenet<->ARPANET connection? Bill Mitchell whm.arizona@rand-relay arizona!whm