[Biochemist]

Transfering on off-topic discussion on CP/M from Digital Research and VAX/VMS from DEC from Social Sciences:

Originally Posted by Qfwfq

The TCP/IP layers are just a simplification of the seven OSI layers APSTNDP. How much real difference is there anyway, between network and data-link? And how much should application presentation and session be the competence of the networking API anyway?

TCP/IP only deals with layers 3 and 4 (TCP is Transport, IP is Network) of the stack. Data link is the lower protocol (e.g., the CSMA/CD message handling structure of ethernet) and the Physical layer is the voltage standards for the signal. OSI surfaced when there were a number of competitive network standards (ethernet, ethernet 802.11, DECNET, Xerox XNS, IBM token ring, Apollo token ring, Ungermann-Bass NetOne, Banyan vines, LanManger, SNA,Appletalk and I think a couple of others) and attempted to reconcile a mass of complexity. It died mostly because ethernet and TCP/IP emerged as the most common standard for the bottom four layers, and the upper layers were not all that important to resolve. (Although Presentation was resolved by defaulting to ASCII over EBCDIC).

For a while, Session control was thought of almost exclusively as a network funciton (as with IBM SNA). With the advent and rapid uptake of the Web (with asynch communication as the standard) the importance of session control within the network declined.

CP/M was no where near as sophisticated as VMS. VMS would even support an on-line monitor and separate background batch processing logic concurrent with foreground OLTP processing. CP/M had no on-line monitor tools and only the most rudimentary batch processing.

[Buffy]

OSI was theory trotted out without any real practice and overly influenced by the initial approaches to networking, like SNA and its forebearers, where IBM always assumed that anything outside the computer room was a teleprocessing monitor concentrating blockmode transactions back to the mainframe, or just mainframes communicating in a trusted network.

The VMS command processor is still totally baroque, and at one of my old companies we used to marvel that the installation scripts on VMS were usually twice as long as on Unix, even with all the multiplatform options on the Unix side. Long live csh!

Tar baby for alexander: The reason Windows is so broken is that it was designed by the same guy who designed VMS....

Cheers,
Buffy

[Biochemist]

Originally Posted by Buffy

OSI was theory trotted out without any real practice and overly influenced by the initial approaches to networking, like SNA and its forebearers, where IBM always assumed that anything outside the computer room was a teleprocessing monitor concentrating blockmode transactions back to the mainframe, or just mainframes communicating in a trusted network.

Well, speaking as a guy that was once in charge of a project to implement communications and graphic display services between about 10,000 workstations when individual subnetworks were running 1)Novell 2) Ungerman-Bass NetOne 3) OS2 EE/LanManager 4) Sun TCP/IP 5)Apollo token Ring and 6) one of the Apple protocols, we would get off work at 10:00Pm and think how nice it would be to have a single common communication stack. That was the environment where OSI surfaced as a value.

The VMS command processor is still totally baroque, and at one of my old companies we used to marvel that the installation scripts on VMS were usually twice as long as on Unix, even with all the multiplatform options on the Unix side.

Maybe, but VMS did a lot of stuff that Unix boxes never did (notably support for VAX Clusters and true pseudoconversational application processing). The more flexibilitiy, the more difficult to implement. That is why IBM MVS hosts are so ugly to set up.

[Buffy]

Originally Posted by Biochemist

...when individual subnetworks were running 1)Novell 2) Ungerman-Bass NetOne 3) OS2 EE/LanManager 4) Sun TCP/IP 5)Apollo token Ring and 6) one of the Apple protocols, we would ... think how nice it would be to have a single common communication stack. That was the environment where OSI surfaced as a value.

Tee hee! No kiddin! I used to work on trying to get databases across all those platforms to talk to each other. I know exactly what you're talking about! As usual though, before the various parties (I was actually one of them), could agree on how it should work, the market decided for us, much to the shagrin of the folks at Apollo, and U/B who had to work with the other guys. Apple and Novell just went proprietary in homogenous envronments with bridges, and luckily for IBM customers, they got smart fast and got people moved with minimal pain.

Originally Posted by Biochemist

...but VMS did a lot of stuff that Unix boxes never did (notably support for VAX Clusters and true pseudoconversational application processing).

Oh no question, I was only talking about the command processor. My first company's product was built on BSD in research and rewritten completely for VMS on the commercial side, and clusters were invaluable in their day for scalability. The cluster code on VMS was also much easier to deal with than the various Unix implementations that finally crept out *years* later. Writing scripts in the command processor though was a total pain in the rear and I do not miss it...

Cheers,
Buffy

[Qfwfq]

Originally Posted by Biochemist

TCP/IP only deals with layers 3 and 4 (TCP is Transport, IP is Network) of the stack.

That's what I most essentially meant! IOW, the 7 OSI layers comprise "TCP" and "IP". Of course TCP/IP is a whole suite, not just transport and network, but we're talkin' from the formal pov. Aren't we?

I may have ranted loosely about the border between N and D but I've never bothered with knowing the borders between Network and below. To me there is IP... and there are the chips on the card, which has to be given its driver's installation and that's where I set the IP address anyway. !!!

I also agree that session should be a matter of the app, liberty to choose; a stateless protocol like pure HTTP or keep connections open day and night and maybe pool them. Of course, some such choices might be conveniently shared between things on a machine but it's got less to do with the actual networking, it needn't be such a fixed standard. Invent fifty dozen of them as needed!

Originally Posted by Biochemist

CP/M was no where near as sophisticated as VMS.

Profuse apologies if I came across as being so highly blasphemous, but I think I said VMS certainly gave more possibilities, certainly more than one could expect on a Z80 machine! I even meant it concerning the humble user, like I was on my university's VAXes.

[Qfwfq]

Originally Posted by Buffy

OSI was theory trotted out without any real practice and...

That's also something like what I meant, I see it like TCP/IP is a "practical" (and simplified) implementation of the OSI model, although I guess there was some kind of implementation for OSI as well.

Originally Posted by Buffy

The VMS command processor is still totally baroque, and at one of my old companies we used to marvel that the installation scripts on VMS were usually twice as long as on Unix, even with all the multiplatform options on the Unix side. Long live csh!

I have always liked hacking and that's why my language is either C or C++. Scripting and typing in commands are, however, two different kettles of fish. In the former it's OK to have a language that is Baroque (sorry, I see it the other way around, nigra sum sed formosa, RICERCAR et quærendo invenietis. Intendami chi vuol, che m'intend'io.) because it's a form of programming and, if your in the mood for it, Baroque things can be fun!

The latter is, in essence, telling the G'ddamned machine what the hell to do, i. e. "just" barking out orders and having them carried out. In this context, it's a drag to consult man pages for most options you need and even for the actual command verbs you have less {rec/frequ}ently used. I guess I just haven't used Unix much but an anti-mnemonic lexicon isn't the best thing for on the fly command lines, they should take a jiffy to learn and get the hang of, without needing constant use, to keep them up there.

I believe in having a clear and mnemonic DCL, all the better if you can then choose your scripting lingo and maybe even devise your own. Yes. I do agree with that about the Posix family!

Originally Posted by Buffy

The reason Windows is so broken is that it was designed by the same guy who designed VMS....

????

Yes, OK, I know NT was developed by a team including some o' the Big Boyz from VMS... but also including some from Unix. The VMS Boyz are from whence NT got it's helium-tight virtual memory and security. NT 4.0 I've heard being considered the most reliable Windoze and 2K has been getting there. I have mainly used these on the job and I don't usually have trouble, excepting some of the feature hitches that have always been around.

I don't know of up to 98 and ME having been designed by a VMS guy.

One of my best friends was administrating VMS systems for years and till the bitter end and also does the various Unix and Windozes. He considered Unix and VMS on a pros'n'cons basis and has always copiously spat upon Uncle Bill. I suspect he would also at those words of yours.

[Buffy]

Originally Posted by Qfwfq

I have always liked hacking and that's why my language is either C or C++. Scripting and typing in commands are, however, two different kettles of fish. In the former it's OK to have a language that is Baroque ... because it's a form of programming and, if your in the mood for it, Baroque things can be fun! The latter is, in essence, telling the G'ddamned machine what the hell to do, i. e. "just" barking out orders and having them carried out.

Oh in the "olden days" you really needed to write scripts to do installations (you could do it C, but because there were lots of useful commands built into the command processor, you'd just have a lot of "exec("copy dev:[install.blort]*.* sys:[appdir.blort]") style calls that make it hard to maintain over time. Go back to my wrench/hammer quote above... I used the term "baroque" in its "obscure (i.e. your "anti-mnemonic"), backwards,innefficient" sense. C/C++ lets you do anything you want of course... but on that cp/shell issue:

Originally Posted by Qfwfq

In this context, it's a drag to consult man pages for most options you need and even for the actual command verbs you have less {rec/frequ}ently used. I guess I just haven't used Unix much but an anti-mnemonic lexicon isn't the best thing for on the fly command lines

Well, you have to understand that a lot of it was economy--early unix versions had short file names, and thus needed short command names too--and a lot of what have become the standard commands on unix were actually written by a bunch of grad students at Berkeley (BSD) with, uh, peculiar senses of humor... My favorite is "biff" (not very useful today, but turned async mail notification on and off in your shell): it was just someone's nickname. Has no meaning whatsoever. "awk" was an interjection. I can go on....

Originally Posted by Qfwfq

Yes, OK, I know NT was developed by a team including some o' the Big Boyz from VMS... but also including some from Unix....

I dropped some characters there: yes I was referring to NT. Yes that's where it got great memory and multiprocess management, but VMS is also where we picked up VMS's horrid DeviceName:[Dir1.dir2]filename syntax. At least they dropped the square brackets, but we unix folks so wished that the microsofties would have seen the light on using a virtual file space and logical links to disks in the file system navigation.

Originally Posted by Qfwfq

One of my best friends was administrating VMS systems for years and till the bitter end and also does the various Unix and Windozes. He considered Unix and VMS on a pros'n'cons basis and has always copiously spat upon Uncle Bill. I suspect he would also at those words of yours.

Ha ha! VMS vs Unix used to be just as religious as Linux vs. Windows. I'm firmly in the *nix camp, even though alexander occasionally accuses me of being apostate...

Cheers,
Buffy

[alexander]

The reason Windows is so broken is that it was designed by the same guy who designed VMS....

that is only one of sooooooooooooooooooooooooooooooooooooooooooooooo.. .................................................. ... many problems it has, if only a month before the dos commercial a Unix commercial aired, none of this mess that we are experiencing with bad microsoft products would be here, hopefully...

Oh in the "olden days" you really needed to write scripts to do installations (you could do it C, but because there were lots of useful commands built into the command processor, you'd just have a lot of "exec("copy dev:[install.blort]*.* sys:[appdir.blort]") style calls that make it hard to maintain over time. Go back to my wrench/hammer quote above...

you poor guys, the old days of no PMS (you have no idea how sorry i am for you)

bunch of grad students at Berkeley (BSD)

look at BSD today, those grad students have made a signifficant investment of their time and devotion to Unix (however bad their code was), ofcourse i think that GNU Unix was quite a way better, they are pretty equal now though, security and stability-wise anyways, BSD's been progrssing in a different direction lately though with their plan to rewrite cisco software and distribute it with their system for free

even though alexander occasionally accuses me of being apostate

Please Buffy, i never in my dreams doubt that you are too knowledgeble of a person to go down the win road, and although i know that sometimes i get too carried away and take my side firmly, i understand that you are just trying to present the other side of the argument, really enjoy talking to all of you for that matter, and you dont have to be a hacker to use Linux, you do not need to run command line, or try to solve problems all the time, I only have problems with people who run windows are refuse to accept that there is another side to the argument and there is something better out there.
BTW however i know that VMS/Unix arguments were as religious as win vs. Linux, i do not however agree that they were as similar of an argument as it may seem...
Oh and i also dont think that OpenVMS is that bad of an implementation of VMS, definately has its users.
Oh, and i think that Clustering under Linux has definately greatly improved, Penguin computing clusters are awesome (mainly because the main system sees your cluster as a single computer and you do not have to have an OS installed on the boxes under your command ), the cluster knoppix clustering is kind of cool because of its bendability and easiness to implement (connect your computer to the network and reboot (netboot), and bang, its a part of the cluster, programming for it is really a great pain in the rear, but its a Beowulf-like cluster)(oh and can anyone explain to me why Beowulf clusters have such dumb specifications (you specifically can not run linux among other thigs)

[Buffy]

Just a couple of clarifications:

There was the BSD up to about 4.3/4.4 that was done at UC Berkeley, and had lots of participants but would not have happened at all without Bill Joy and a handful of others. When BSD systems was created as a separate company, basically none of these people was involved, and what has happened since has been a commercial endevour.

Etymology of the word Open: in the late 80s and early 90s "Open" meant *only* standards compliant. Unix flavors all had to comply with POSIX, OSF, or other candidate standards that popped up all over the place. It had NOTHING to do with "open source" as almost all of these variants were based on original licensed AT&T code, and *could not* be even shown to anyone who had not signed a license agreement. But they could have the standard library interfaces that would allow applications to be portable across systems simply by recompiling. Or at least that was the theory. Those of us who had low-level applications had to make accomodations to underlying architectures (byteswapped?, page sizes? raw partition calls?) even with the standards. And then of course most of the vendors did the standard libraries as an afterthought and optimized their native libraries with stuff that all app vendors wanted/had to use, so it all degraded into anarchy, opening the door for Windows NT to start making inroads as a low-end server platform. Oh well.

Why do I mention this stuff? Because you referred to "OpenVMS" which was simply DEC adding an OSF and Posix compliant library into the standard VMS distribution. It caused much laughter in the community at the time because it wasn't as if it made it possible to take your Unix apps and recompile them unchanged on VMS, but the name-change was permanent. There's no difference between VMS and OpenVMS: the latter is just the name on the versions after about 1993....

Cheers,
Buffy

[Qfwfq]

Cheers,

Originally Posted by Buffy

I used the term "baroque" in its "obscure (i.e. your "anti-mnemonic"), backwards,innefficient" sense.

I reckoned that, but it applies more to *ix than to:

Originally Posted by Buffy

The VMS command processor is still totally baroque.

Besides, I have nothing against Baroque. I'm not so hot on Baroque architecture but those composers were great! I just slipped a CD of Monteverdi into the PC here, with Kirkby, Thomas and Partridge on the vocals. Not backward, not inefficient.
Yesterday I forgot to say, I find bash better than csh.

Originally Posted by Buffy

Well, you have to understand that a lot of it was economy--early unix versions had short file names, and thus needed short command names too

I was quite accustomed to short names, in the days of hacking on Z80 in a few kBs I was good at packing stuff into few bytes and I never lost the habit quite totally. But I remember even VMS having 8+3 filenames! Nonetheless, directory is longer than 8. As for quick typing you didn't even need completion, once you resolved the ambiguity you could leave it at dir, dire, direc, direct... and that was just that. According to how much in practice you were, you could type more to be sure, or less. If you slipped, the later versions made it quick to clarify. Very comfortable for typing in command after command. Option and params weren't cryptic like -d -dr -D etc. Redirection? out=sys:[appdir.blort]foo.bar!!!

Much as *ix path syntax and features are good, I didn't find the VMS syntax unworkable, far less horrid.

I agree with using sense of humour in choosing names, a bit less in command syntax, first thing that the dumb user has to get immediately familiar with. When programming I like choosing identifiers whimsically. One example of mine comes to mind, I had to do part of a gui, a Swing control that would view the local file system and the FTP remote one. At first customer specs were incomplete so I quickly worked it so that, if the control was instanced with a null pathname in the constructor it would show all drives, and allow the user to choose a root one and even change it. This meant have a root node that was nothing on the file system and thus hidden, and I needed to use the pointer here and there to accordingly change behavoiur of things. I wanted a name that would be suggestive to myself but not too long, thinking of all aspects including that the user would mustn't ever see it I thought "Horatio" and used that.

Originally Posted by Buffy

Ha ha! VMS vs Unix used to be just as religious as Linux vs. Windows. I'm firmly in the *nix camp, even though alexander occasionally accuses me of being apostate...

My old friend is a staunch atheist, all the same he is very fond of penguins. He was also very enthusiastic in his VMS days and still considers it was a great OS, despite a few *ix advantages.

Originally Posted by Buffy

Etymology of the word Open: in the late 80s and early 90s "Open" meant *only* standards compliant. Unix flavors all had to comply with POSIX, OSF, or other candidate standards that popped up all over the place. It had NOTHING to do with "open source" as almost all of these variants were based on original licensed AT&T code, and *could not* be even shown to anyone who had not signed a license agreement........

Agreed. I pointed out the diff to Bio on the fiber of the other thread, that became this thread. ¡W OSF! ¡~W Uncle Bill!