pag@hao.UUCP (Peter Gross) (08/19/83)
I tend to agree with Lauren that charging for optimal path information is not the best way to do things. However, the task of compiling and maintaining that kind of information is by no means trivial. Whoever does this job definitely deserves renumeration for their time and effort. Perhaps there might be some way for USENIX (ie, as opposed to USENET) to pay for this service for its members. I see a similarity to the "official" recording of conference notes and proceedings. In the old days, this was done on a volunteer basis, and generally you got what you paid for. Now that this is done professionally, the quality is much higher (though 'twould be nicer to get the proceedings a bit sooner). Couldn't we do the same with mail routing information? I realize that not all members of USENIX are members of USENET, or vice-versa, but it seems like a fairly reasonable approach. --peter
goutal@decvax.UUCP (Kenneth G. "Kenn" Goutal) (08/24/83)
I can agree that compiling path info is by no means trivial, and that if you're doing it by hand it's a pain in the neck. But don't we already have programs that do this? If not, why not? I feel constrained to point out that DECnet (which is set of protocols and code implementing them, not a specific set of wires and machines) does this automatically -- it's just a part of normal protocol to pass routing information around. Now, I realize that actually changing uucp so that it did this could be a collossal task, and will probably never happen. However, it does not seem unreasonable to suggest that path-building software could somehow be grafted on top of Usenet, based on existing software and conventions. -- Kenn (decvax!goutal)
smb@ulysses.UUCP (08/25/83)
The hard part isn't calculating optimum paths, it's collecting good data. I know; I've tried it. First, most sites do not respond to requests for info. Those that do typically ignore the suggested format, meaning that you have to do a lot of editing. There are also conflicts to be resolved. For example, some university will say that some site polls them hourly, but the polling site denies that it's more than daily. New sites come on, old ones drop off. Automated schemes depend on co-operation -- will all sites install the program you write? Probably not. Even if they do, what's their source of data? L.sys files are notoriously full of non-host info. For example, some sites have other time-sharing machines listed, and a 'cu' variant that dials up the machine, logs them in, etc. I once wrote an automated line loopback tester -- but I used a lot of uucp code, including L.sys, to establish the link and fire up the remote end (for cases when one couldn't put the line into loopback mode via hardware). Then there are the sites that don't forward mail, or that use something other than '!'... --Steve