On 2013-12-20 00:35, Clem Cole wrote: I remember having this discussion about 30 years ago with a couple of the VMS implementors. Im not going to argue with you - other thanks to say - at the end of the day we decided just what I said -- there are pros/cons to each solution - you pointed out some of the pros for logicals. It's an endless discussion, and one I hoped we would not need to have once more. It's pretty pointless. I can't do with environment variables what I'm now doing with logical names for my resolver in RSX. It's sad (and boring) when people try to convince me that I'm wrong. I know very well how it all works. As for a universal resolver vs having each client do a lot of work in the library, that was and is an implementation choice - left over from how it was the original DNS was implemented (at Berkeley on the VAX at the time). It is interesting that the UNIX idea of a small program was sort of lost by many of my brethren at UCB when large address spaces showed up. I suspect that if the DNS had been written for PDP-11 UNIX it might have been a separate process. At which point, environment variables would not have been possible anyway, since they are always local to the process. The fact that the resolver works the way it do is indeed just a choice, but I can't see what other choice that would have made sense in that world. Doing it in a separate process is what the DNS server gives you. So it's not that this has not been done. It's just a question of what makes sense and is convenient. Johnny On Thu, Dec 19, 2013 at 6:18 PM, Johnny Billquist <bqt at softjar.se <mailto:bqt at softjar.se>> wrote: Hi. On 2013-12-20 00:07, Clem Cole wrote: logical names have a unix equiv. in environment variables. Sortof, but not completely. there are some differences and advantages to both but the idea and practice ends up being the same IMO I disagree. what you are doing in resolve$foo could just as easily be put in the environment and get the same result No. Because I have (obviously) system wide logical names, which provides the system defaults for this. Any user defined logical names will override the system ones. This concept of having logical names on several levels (system, group, login, session, task) do not have an equivalent form in Unix. There are more aspects of environment variables that differs from logical names as well. Which also is a part of the reason you don't use environment variables for some things, and instead do have the small text files. (And I love small text files, by the way, so don't misunderstand me.) It's just that in some situations, other solutions can give a different bias, which might make some things easier. that said you are correct that unix often used small txt files for some set up and configuration but some programs used the env too. truth is sna files were used in some of the dec OSes too - it was just a matter of taste. I don't think one is better than the other -- I do think when in Rome I would want to use the scheme that is the operational standard under the least astonishment principle I'm not even going to get into an argument about which is better, or which thing was done in different places. That is a totally different question. Both ways work, they provide different pros and cons. I just totally love the solution I have come up with for name resolving in RSX, which allows me to easily solve some details which I have found very hard to do in a simple, clean and nice way with the resolver solution that exists under Unix. Johnny Clem On Dec 19, 2013, at 5:14 PM, Johnny Billquist <bqt at softjar.se <mailto:bqt at softjar.se>> wrote: On 2013-12-19 21:07, John Wilson wrote: From: Johnny Billquist <bqt at softjar.se <mailto:bqt at softjar.se>> To get more on topic - I'm working on a name resolver for RSX right now. It's a totally awesome thing, and I only wish Unix (or even VMS) had something similar. Me too! Port? To Unix? Not likely. Logical names just don't have a good equivalent. The Unix way is text files with configurations instead. But the form of that is already established, with the /etc/resolv.conf and /etc/nsswitch.conf (and /etc/hosts) files. Hard to change, and those do unfortunately not easily allow user overrides at any level. (Now I just need to finish the DNS part of it.) Details. Sounds like good engineering so far. Nice going! Yeah. The DNS part is not really difficult. It's just a question of setting up and sending out UDP packets, and then parsing the responses. Although I must admit that DNS have some bits that I'm not too happy about. Working on building the DNS query right now. That is actually really simple. Doing the parsing will probably take a few days of work. Johnny -- Johnny Billquist || "I'm on a bus || on a psychedelic trip email: bqt at softjar.se <mailto:bqt at softjar.se> || Reading murder books pdp is alive! || tryin' to stay hip" - B. Idol -- Johnny Billquist || "I'm on a bus || on a psychedelic trip email: bqt at softjar.se <mailto:bqt at softjar.se> || Reading murder books pdp is alive! || tryin' to stay hip" - B. Idol -- Johnny Billquist || "I'm on a bus || on a psychedelic trip email: bqt at softjar.se || Reading murder books pdp is alive! || tryin' to stay hip" - B. Idol