barry@zaphod.uchicago.edu (Barry Merriman) (08/06/89)
-
In previous posts, I have seen it suggested that NeXT, by
offering a $6500 machine directed at students, may be
trying to return us to the era of centralized computing, with
(small) NeXT clusters provided for (large) student bodies.
I have also seen it suggested that NeXT should price in at $4k,
to compete with low end SPARCstaions.
As a member of the target user group (student---4 years undergrad,
5 years grad, just getting PhD.), I have the following comments:
(1) For the majority of students, centralized computing was, is
and will continue to be a necessity---because of (lack of) money.
In my 9 years, I have never had enough money to get even the cheapest
IBM clone, much less a Mac, Mac II, SPARC, or NeXT. Even If someone
_gave_ me a NeXT with a NeXT Laser Printer, I would have a hard time
buying the $120 toner cartridges for it---I'm poor, and so are
many of my friends. We can't afford the NeXT if they give them away!
(2) The only computing available has been centralized---I'm just
glad that we got away from the "mainframe" concept
(undergrad years: 1 CDC and 2 Vaxen [with ~20 users _at a time_]
and operating systems from hell, for 30,000 students---10,000:1 ratio),
and advanced to the "cluster of graphic workstations" concept
(grad years: 3 Sun 3/60's for the 80 students in the Math Dept.
and 40 Macs and 16 Mac II's for the 8,000 students in the U of Chicago
---27:1 and 175:1 ratios).
NeXT _has_ planned an advance of this current state of affairs.
As I was reading through the NeXT propaganda, fantasizing, I came
across their vision of the (immediate) future. The idea is for
a University to provide a NeXT cluster, but for each student
user to provide their own $55 optical disk (I _can_ afford that)
which, having 256MB, carries the users entire _computing world_
---operating system, applications, data, etc. In this way, each user
can totally personalize the machine, as if it were their own. And
they can transparently and effortlessly compute at different clusters,
even at different universities. Your own computer (environment) for $55
---This is a further step towards the ultimate
decentralization of personal computer ownership.
It would sure beat the hell of of a Mac II cluster---unless your
world is small enough to fit on a 1.4 MB diskette %-).
Well, in short, all most students can afford is to use a University
cluster, and I'd rather use a NeXT cluster than just about any other
(save, perhaps, a Sun cluster, but the interface/environment is not
as pleasant---certainly for people shy of UNIX, as most college students
are inclined to be.)
Any comments...?
-Barry Merriman
Grad Student (for three more weeks) That said, I'll be buyin'
University of Chicago, Dept of Math my own Cube as soon as I
get a few real paychecks...jgreely@oz.cis.ohio-state.edu (J Greely) (08/08/89)
In article <4866@tank.uchicago.edu> barry@zaphod.uchicago.edu (Barry Merriman) writes: >The idea is for a University to provide a NeXT cluster, but for each >student user to provide their own $55 optical disk (I _can_ afford >that) which, having 256MB, carries the users entire _computing world_ >---operating system, applications, data, etc. In this way, each user >can totally personalize the machine, as if it were their own. ...and while I'm dreaming, I want a pony. :-) 256 meg makes a nice data disk, but it is sub-optimal as a system disk. The first problem I have with the "world in a pocket" idea is that those NeXTs will of necessity be standalone or minimally networked. The second problem is space: a bootable optical disk under 0.9 has barely 18 meg free, nowhere near enough for my "computing world". I can trim that disk down quite a bit, but it still won't have enough space free to be very useful to me (nor will that stripped-down disk be easy for the average user to create, although BuildDisk should cut down on the effort required). If I were putting together a cluster of NeXTs here, they would be diskless clients of a non-NeXT server, hooked into the department network. A few standalone machines for casual use wouldn't be a bad idea, but they would not be networked to the rest. >And they can transparently and effortlessly compute at different >clusters, even at different universities. Your own computer >(environment) for $55 ---This is a further step towards the ultimate >decentralization of personal computer ownership. You're right, you *have* been reading propaganda :-). It's an interesting dream, but it still needs some polish. In five years, that model may be practical, but I don't think either NeXT or the university market is ready to support it successfully now. I wouldn't be too unhappy if I were proven wrong, though. >It would sure beat the hell of of a Mac II cluster---unless your >world is small enough to fit on a 1.4 MB diskette %-). My world fits on a 256 Meg disk, but it would be more comfortable on two. Between source code, graphics, and sound, I chew up lots of space. >That said, I'll be buyin' my own Cube as soon as I get a few real >paychecks... Have a blast. The few times I've taken mine home for the weekend, it's made a nice PC. A bit loud, but nice. If I had the $13,000 lying loose, I'd consider picking one up myself. -=- J Greely (jgreely@cis.ohio-state.edu; osu-cis!jgreely)
dorner@pequod.cso.uiuc.edu (Steve Dorner) (08/08/89)
In article <JGREELY.89Aug7190716@oz.cis.ohio-state.edu> J Greely >... > 256 meg makes a nice data disk, but it is sub-optimal as a system >disk... > ... The second problem is space: a bootable optical disk under >0.9 has barely 18 meg free, nowhere near enough for my "computing >world"... > > If I were putting together a cluster of NeXTs here, they would be >diskless clients of a non-NeXT server, hooked into the department >network. A few standalone machines for casual use wouldn't be a bad >idea, but they would not be networked to the rest... There is a lot here to agree with. I don't think allowing students to boot from their own optical disks is a good idea at all, at least unless and until NeXT does something Kerberros-like, so that security is not a problem. On the other hand, an optical disk AS A DATA DISK has a heck of a lot to recommend it. We sure won't give students 200MB of disk space on any of our machines; I know at least one person who paid $50 out of his pocket, just to have some disk space for his music research. Finally, let me comment that, with a little surgery, it's easy to reclaim 100MB or so from most NeXT disks, if they're networked. Just nuke /NextLibrary, and mount one from another machine. The stuff in /NextLibrary is accessed only intermittently, and sharing it over the network is not a performance hit on network, client or server (unless of course the server is down, which is very bad news indeed). I've been doing this for some time now, and really enjoy having the extra space. -- Steve Dorner, U of Illinois Computing Services Office Internet: s-dorner@uiuc.edu UUCP: {convex,uunet}!uiucuxc!dorner IfUMust: (217) 244-1765
greid@adobe.com (Glenn Reid) (08/09/89)
In article <JGREELY.89Aug7190716@oz.cis.ohio-state.edu> J Greely <jgreely@cis.ohio-state.edu> writes: >...and while I'm dreaming, I want a pony. :-) > > The first problem I have with the "world in a pocket" idea is >that those NeXTs will of necessity be standalone or minimally >networked. > If I were putting together a cluster of NeXTs here, they would be >diskless clients of a non-NeXT server, hooked into the department >network. A few standalone machines for casual use wouldn't be a bad >idea, but they would not be networked to the rest. I presume the reason you would not network these machines together is "security"? Security can be a false idol, I think. If you pretend the NeXT machine is yet another Vax or Sun or Apollo UNIX machine, and you tie it into your existing network, then you have some security problems to worry about, mainly because you are operating under the premise that your network is already secure. One of the main reasons that UNIX "security" is such a big issue is that there has been, traditionally, almost NO way for users to take data off the system and put it in their pockets. Professors are forced to leave exams and grades on the system, etc. What else would you do? Write out a tape and then read it back in when you want to work? If you look at microcomputers, nobody even has a password on many of the systems. If something is sensitive, you put it on a floppy and lock it up in your desk. With the NeXT optical disks, that becomes completely practical, and the notion of security is less an issue, I think. >You're right, you *have* been reading propaganda :-). It's an >interesting dream, but it still needs some polish. In five years, >that model may be practical, but I don't think either NeXT or the >university market is ready to support it successfully now. I wouldn't >be too unhappy if I were proven wrong, though. If the optical disk were, say, three times as fast as it is now, don't you think this model would be practical today?
jgreely@oz.cis.ohio-state.edu (J Greely) (08/09/89)
In article <1054@adobe.UUCP> greid@adobe.com (Glenn Reid) writes: >In article <JGREELY.89Aug7190716@oz.cis.ohio-state.edu> J Greely > <jgreely@cis.ohio-state.edu> writes: >> The first problem I have with the "world in a pocket" idea is >>that those NeXTs will of necessity be standalone or minimally >>networked. >I presume the reason you would not network these machines together is >"security"? It'll do for the word, although "integrity" and "authentication" are strong contenders. >Security can be a false idol, I think. If you pretend the NeXT machine >is yet another Vax or Sun or Apollo UNIX machine, and you tie it into >your existing network, then you have some security problems to worry >about, mainly because you are operating under the premise that your >network is already secure. I have no illusions about the security of our network, but I do know that adding a machine to it that's bootable by J Random Undergrad off of his own disk will drastically lessen what "security" we *do* have. For a quick example of the headaches, consider: the mythical personal boot disks will either have 1) random uids and usernames, or 2) all the same. Neither of these is terribly useful if you want to use network services such as NFS and shared printers. >One of the main reasons that UNIX "security" is such a big issue is >that there has been, traditionally, almost NO way for users to take >data off the system and put it in their pockets. Professors are forced >to leave exams and grades on the system, etc. ... >If you look at microcomputers, nobody even has a password on many of >the systems. If something is sensitive, you put it on a floppy and >lock it up in your desk. With the NeXT optical disks, that becomes >completely practical But only if everyone has constant access to a NeXT! In the world as it is *now*, there are usually large numbers of "foreign" machines, like Suns, Vaxen, and Pyramids, and they aren't going to vanish anytime soon. I can't write to an optical from those, so unless I can physically get to a NeXT whenever I need to read or write sensitive data, I've gained nothing. In the unlikely case that an installation has only NeXTs, your view will be practical. Until and unless that becomes commonplace, the NeXT will have to be no less "secure" than other networkable Unix boxes (and it is supposed to be, under 1.0). I'll just mention the idea of sensitive information that must be accessed by more than one person, such as proprietary source code under development. Not something that can be conveniently dumped to OD and locked in a desk, unless the entire development team shares an office. >>In five years, that model may be practical, but I don't think either >>NeXT or the university market is ready to support it successfully now. >>I wouldn't be too unhappy if I were proven wrong, though. >If the optical disk were, say, three times as fast as it is now, don't >you think this model would be practical today? If it were available at the same price (or "reasonable"), for a wide variety of hardware, yes. If it can only be read on a NeXT, no. Currently, I have an OD that's well-filled with personal archives. When I leave the university (which may be quite soon), I'll be transferring it all to magtape, because I can read it anywhere. The OD is more convenient while I have a NeXT on my desk, but once that's gone... -=- J Greely (jgreely@cis.ohio-state.edu; osu-cis!jgreely)
crum@lipari.usc.edu (Gary L. Crum) (08/09/89)
Constrained and controlled resources may be sufficient for given tasks, but luxury and freedom contribute to human inspiration and creativity. Let's find solutions to bring the stimulation of UNIX system control to more than administrators and owners of systems -- to users of student workstation clusters as well. If you manage school systems and have something like this ideal as a goal, please advertise it. You might help someone choose the best environment. Gary (doctor of Philosophy, someday)
deke@ee.rochester.edu (Dikran Kassabian) (08/09/89)
In article <1054@adobe.UUCP> greid@adobe.COM (Glenn Reid) writes: < In article <JGREELY.89Aug7190716@oz.cis.ohio-state.edu> J Greely <jgreely@cis.ohio-state.edu> writes: < < > The first problem I have with the "world in a pocket" idea is < >that those NeXTs will of necessity be standalone or minimally < >networked. < > If I were putting together a cluster of NeXTs here, they would be < >diskless clients of a non-NeXT server, hooked into the department < >network. A few standalone machines for casual use wouldn't be a bad < >idea, but they would not be networked to the rest. < < I presume the reason you would not network these machines together is < "security"? I feel like its time to add my $.02 here. Security is indeed one con- sideration here, but so is 'correct' network behaviour. Not everyone is network connected to the arpanet, but those who are have additional concerns. I personally would worry that a student might wish to boot up one of our network connected NeXT boxes with his/her own boot OD. If things on that boot OD did not conform to the way in which my domain operates, I may have a problem. Examples? Declaring some non fully qualified domain name and then broadcasting beyond the confines of our domain. This is antisocial, and people like me are responsibile for seeing that this doesn't happen. How about when the student boots up with version X of some networking or printing software when the rest of us in my domain are using version Y. Perhaps something else will break. So I'm not only worrying about intentional tampering, I'm also concerned that inadvertant misconfiguration will have adverse effects that are more widespread than within one machine. < Security can be a false idol, I think. If you pretend the NeXT machine < is yet another Vax or Sun or Apollo UNIX machine, and you tie it into < your existing network, then you have some security problems to worry < about, mainly because you are operating under the premise that your < network is already secure. Which, of course, many of us do not. Security is relative. I don't give out keys to the machine room or root passwd, but I don't kid anybody either. When I was interviewed for my present position three years ago, my boss asked me how to make a computer secure. I said "remove all network and serial connections, lock it in a room, throw away the key, and then it will be almost secure." < If you look at microcomputers, nobody even has a password on many of < the systems. If something is sensitive, you put it on a floppy and < lock it up in your desk. With the NeXT optical disks, that becomes < completely practical, and the notion of security is less an issue, < I think. I love the idea of using the OD for data. Cheap, portable, versatile. I just want to be careful that we don't fall into potentially cavernous pitfalls. ^Deke Kassabian, deke@ee.rochester.edu or ur-valhalla!deke Univ of Rochester, Dept of EE, Rochester, NY 14627 (+1 716-275-3106)
jgreely@oz.cis.ohio-state.edu (J Greely) (08/14/89)
In article <CRUM.89Aug9024441@lipari.usc.edu> crum@lipari.usc.edu (Gary L. Crum) writes: >Constrained and controlled resources may be sufficient for given tasks, >but luxury and freedom contribute to human inspiration and creativity. Y'know, I've been putting off replying to this for days, mostly because I haven't the slightest idea what you're talking about. It's obvious, I think, that you have adopted the artist's model of a computer as a creative tool, and I get the impression that what I think of as system administration is what you think of as fascism. Then again, I *could* be wrong. >Let's find solutions to bring the stimulation of UNIX system control >to more than administrators and owners of systems -- to users of >student workstation clusters as well. Control? I'd appreciate a clarification of what you're saying here, because my first impression is that you're sponsoring me for an ulcer. What exactly does that interesting word "control" mean here? >If you manage school systems and have something like this ideal as a >goal, please advertise it. You might help someone choose the best >environment. What ideal? Perhaps you should expand on the environment you think is "best", so we can tell you how close it is to what people are doing. Some people with experience in creating and administering such environments might have a few bones to pick, but the ensuing discussion should at least clear the air on how things *can* be done. "Obedience to the Law is Freedom" -=- J Greely (jgreely@cis.ohio-state.edu; osu-cis!jgreely)