danq@jetson.UUCP (Daniel Quinlan) (03/14/91)
I've been reading the discussion about pc nfs's deficiencies with some
interest. Here at CHS we have a network of Sun workstations, and around
100 completely almost completely unconnected pc's. Obviously, this
is undesirable.
We would like to interconnect all the pc's and the
sun system, and perhaps give terminals to those few other people
who don't already have pc's, terminals, or workstations. Then we could
have an effective email service -- this is really the driving force.
We've been looking into pc-nfs a little bit, and were hoping that we
would get the benefit of shared files (through an nfs-mounted partition
on the sun's), automated backups (at least on the nfs-mounted partitions,
and perhaps from disks on pcs also), and shared printers (currently
many pc's have their own dedicated laser-jet's.)
Our initial impression is that pc-nfs is a bit raw for the average
pc user. We recently looked at a package Great Western is trying
to market to put menu's on top of it. I don't think we can afford
to go that route, and I'm afraid the relative lack of standards in
the pc world for networks would doom us to disappointment with the
approach anyway.
On the other hand, we might be happy just configuring the machines
to automatically mount a few partitions at boot, and to always print
to a specific printer (this is at least no worse than the current
situation). However, I'd like to at least find a mail program which
could read the mail from an nfs mounted /usr/spool/mail, and send
mail via smtp direct from the dos command line. I realize that
we could use the emulator and log into a sun machine and read mail,
but that's a bit too complex for many people here, and it's
certainly not as nice as a local mail program.
So, my questions are:
a) Is anyone using pc-nfs in a situation similar to ours and happy with it?
b) Has anyone found an inexpensive (or free) mail program which would
work on an nfs-mounted /usr/spool/mail?
c) Are there any unobvious concerns I should have about adding
100 pc's to an ethernet network with pcnfs? (I was thinking of putting
them on a separate subnet to isolate them a bit).
d) Given the estimated $30-60k cost of networking 100 pc's, are there
other options which would be better in the long term? Any such option
must at least provide an effective interface for e-mail between the
sun network and the pc network. I am wary of having two different
networks.
Please reply to me, and I'll summarize. Thanks in advance.
danq
--
Daniel Quinlan {uunet,boulder}!chs!danq
System Administrator chs!danq@boulder.colorado.edu
Consumer Health Services 303/442-1111 x3124
5720 Flatirons Parkway Boulder, CO 80301phil@brahms.amd.com (Phil Ngai) (03/14/91)
In article <123@jetson.UUCP> danq@jetson.UUCP (Daniel Quinlan) writes: |c) Are there any unobvious concerns I should have about adding |100 pc's to an ethernet network with pcnfs? (I was thinking of putting |them on a separate subnet to isolate them a bit). If you give 100 users a PC-NFS package and one page of instructions, almost certainly at least one of them will give his PC the same name as the server and crash the server. If you do all the installs yourself this is much less likely to happen. (what a bother) Stupid users. Bad users. Wish they'd leave my nice Sun alone instead of causing trouble. They seem to want to get work done instead of just admiring how wonderful a Sun is. -- The government is not your mother. The government doesn't love you.
eric@mks.mks.com (Eric Gisin) (03/15/91)
In article <1991Mar14.051003.6212@amd.com> phil@brahms.amd.com (Phil Ngai) writes:
. If you give 100 users a PC-NFS package and one page of instructions,
. almost certainly at least one of them will give his PC the same
. name as the server and crash the server. If you do all the installs
. yourself this is much less likely to happen. (what a bother)
Do you give a typical user an ethernet card or a hard disk and say
"here, install this"? Of course not. Nor do you give them
some TCP/IP-based network software with things like IP addresses,
subnet masks, YP, etc, and expect them to be able to configure it.
Why do people expect all software to be as easy to install as
a simple "insert the disk and run" DOS application?
. Stupid users. Bad users. Wish they'd leave my nice Sun alone instead
. of causing trouble. They seem to want to get work done instead of
. just admiring how wonderful a Sun is.TOMIII@MTUS5.BITNET (Thomas Dwyer III) (03/16/91)
>If you give 100 users a PC-NFS package and one page of instructions, >almost certainly at least one of them will give his PC the same >name as the server and crash the server. If you do all the installs >yourself this is much less likely to happen. (what a bother) Ever hear of RARP? We currently have about 300 machines running PC-NFS, and never had a problem with duplicate host names. The only thing the machines need to have unique are the Ethernet addresses, which I'm sure is never a problem with anyone. :-) Cheers, Thomas Dwyer III Email: tomiii@mtu.edu Network Programmer tomiii@mtus5.BITNET Computing Technology Services Voice: (906) 487-2110 Michigan Technological University Fax: (906) 487-2787
sob@tmc.edu (Stan Barber) (03/16/91)
In article <1991Mar14.051003.6212@amd.com> phil@brahms.amd.com (Phil Ngai) writes: >If you give 100 users a PC-NFS package and one page of instructions, >almost certainly at least one of them will give his PC the same >name as the server and crash the server. If you do all the installs >yourself this is much less likely to happen. (what a bother) > >Stupid users. Bad users. Wish they'd leave my nice Sun alone instead >of causing trouble. They seem to want to get work done instead of >just admiring how wonderful a Sun is. We have over 100 users of PC-NFS and we don't have this problem. Perhaps the reason we don't have it is because we have PC people that install the software for the users. I guess this is a really novel (not NOVELL) concept. It works really well for us. It's not really a bother. This way we are able to know who gets upgrades when the next releases come out. We handle all the software we support this way. An aside: Since Phill started this thread of articles following information about PC-NFS 3.5, I wonder if his comments are based on experience with PC-NFS 3.5. If they are, they have serious implications for me. If they are not, they are mostly irrelevant to me. -- Stan internet: sob@bcm.tmc.edu Director, Networking Olan uucp: {rutgers,mailrus}!bcm!sob and Systems Support Barber Opinions expressed are only mine. Baylor College of Medicine
leoh@hardy.hdw.csd.harris.com (Leo Hinds) (03/18/91)
In article <4750@gazette.bcm.tmc.edu> sob@tmc.edu (Stan Barber) writes: >An aside: Since Phil started this thread of articles following information >about PC-NFS 3.5, I wonder if his comments are based on experience with >PC-NFS 3.5. If they are, they have serious implications for me. If they are >not, they are mostly irrelevant to me. I have also asked the same question, but received no answers. Considering the fact that Phil started this (typed) verbal diarrhea right after the announcement, but MORE IMPORTANTLY BEFORE SUN started shipping the product, I would have to say no, he has not seen the product. Now if he (or others @ AMD) were beta sites (my guess is not for long after this), he might have seen it then. On a related side note ... what is the command (for RN) to automatically junk all articles from a specific poster? ... Now that Phil is repeating (for the n^99 time) the SAME STUFF, I would like to follow HIS advice and not read any more of his stuff. leoh@hdw.csd.harris.com Leo Hinds (305)973-5229 Gfx ... gfx ... :-) whfg orpnhfr V "ebg"grq zl fvtangher svyr lbh guvax V nz n creireg ?!!!!!!? ... znlor arkg gvzr
phil@brahms.amd.com (Phil Ngai) (03/19/91)
In article <4750@gazette.bcm.tmc.edu> sob@tmc.edu (Stan Barber) writes: |We have over 100 users of PC-NFS and we don't have this problem. Perhaps |the reason we don't have it is because we have PC people that install the |software for the users. Then you have the added expense of people to install PC-NFS. Believe it or not, Novell IS easy enough to install at the client level that most users can do it themselves. |I guess this is a really novel (not NOVELL) concept. It works really well No, it's not novel, just expensive. If you are rich enough to have it, good for you. Since budget is a concern for most people, I felt it appropriate to point out this additional expense associated with Sun PC-NFS. |An aside: Since Phill started this thread of articles following information |about PC-NFS 3.5, I wonder if his comments are based on experience with |PC-NFS 3.5. If they are, they have serious implications for me. If they are |not, they are mostly irrelevant to me. My comments are based on a careful reading of the Sun Press Releases and magazine articles, some of which actually have pictures. If I have made any errors of fact, I welcome corrections from Sun, whom I believe to read this group. Which issues have serious implications for you? Since you have around 100 Sun PC-NFS installations, what have I said that is news for you? -- The government is not your mother. The government doesn't love you.
phil@brahms.amd.com (Phil Ngai) (03/19/91)
In article <2660@travis.csd.harris.com> leoh@hardy.hdw.csd.harris.com (Leo Hinds) writes: |I have also asked the same question, but received no answers. Considering Sorry if you haven't seen a response. I try to respond to most questions but the cross-posting and followups can be very twisted. |the fact that Phil started this (typed) verbal diarrhea right after the |announcement, but MORE IMPORTANTLY BEFORE SUN started shipping the product, |I would have to say no, he has not seen the product. Now if he (or others |@ AMD) were beta sites (my guess is not for long after this), he might |have seen it then. No, I have not seen the product or any beta site software. I believe AMD is a beta site, but I personally am not a beta software user. My observations are based on the press releases and magazine articles on Sun PC-NFS 3.5. Corrections welcome. -- The government is not your mother. The government doesn't love you.
sob@tmc.edu (Stan Barber) (03/19/91)
In article <1991Mar18.192417.3970@amd.com> phil@brahms.amd.com (Phil Ngai) writes: >My comments are based on a careful reading of the Sun Press Releases >and magazine articles, some of which actually have pictures. If I have >made any errors of fact, I welcome corrections from Sun, whom I believe >to read this group. In other words, your comments relate to your experience with the 3.0.1 or earlier release. That release was available before Windows 3.0 was available to developers, let alone users. This means your comments about Window compatibility are based on the press release (and the pictures) alone. No real user experience....sigh. >Which issues have serious implications for you? Since you have around >100 Sun PC-NFS installations, what have I said that is news for you? Since your opinions are based on use of the version if I have (which we feel works pretty well with Windows given what I just wrote), nothing you have said is news to me. -- Stan internet: sob@bcm.tmc.edu Director, Networking Olan uucp: {rutgers,mailrus}!bcm!sob and Systems Support Barber Opinions expressed are only mine. Baylor College of Medicine
geoff@hinode.East.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top) (03/19/91)
I'm staying out of this flamefest, but I want to comment on one
observation by Leo Hinds (in <2660@travis.csd.harris.com>):
# <answering Stan's question about whether Phil saw 3.5>
#I have also asked the same question, but received no answers. Considering
#the fact that Phil started this (typed) verbal diarrhea right after the
#announcement, but MORE IMPORTANTLY BEFORE SUN started shipping the product,
#I would have to say no, he has not seen the product. Now if he (or others
#@ AMD) were beta sites (my guess is not for long after this), he might
===================================
#have seen it then.
It is not our policy to comment on who our beta customers are. However
I would like to emphasize that AMD is a long-standing customer of ours,
that we value their business, and that (IMHO) Phil is not particularly
representative of the AMD people we deal with. (He said as much in once
of his first postings in this thread.) We wouldn't allow anything like
this, um, "debate", to get in the way of future beta or similar
agreements with AMD or anyone else.
-- Geoff Arnold, PC-NFS architect, Sun Microsystems. (geoff@East.Sun.COM) --
------------------------------------------------------------------------------
-- Sun Microsystems PC Distributed Systems ... --
-- ... soon to be a part of SunTech (stay tuned for details) --barryf@aix01.aix.rpi.edu (Barry B. Floyd) (03/20/91)
Just to add fuel to the fire, I will herein provide information presented in the "newWare Users Internation" NetWare Connection (publication) dated March/April 1991: - on page 5; there is mention of LAN WorkPlace for DOS, Macintosh and OS/2. with little or not technical information. - on page 4; they discuss "IP tunneling", which according the the text of the article, "... the NetWare v3.11 file server envelop(es) IPX packets in a form acceptable to the IP routers on the internetwork." The accompanying example leads me to believe that this feature is only of interest to Netware LAN nodes attempting to exchange IPX packets over the internet. A non-Netware server/client attempting to "read" the packet would get as far as the IP information and then be lost trying to "read" the IPX date within. - on page 5; they discuss NetWare v3.11 "file service enhancements" in the form of "support for a variety of clients, allowing them to use the Netware 3.11 server to store and retrieve files (see figure 2.0)". Apparently NetWare 3.11 now recognizes OS/2's HPFS, Unix's NFS and Mac's OS as well as OSI GOSSIP's FTAM (a subset of NFS support). The figure 2.0 mentioned above clearly shows a NetWare 3.11 file server attached to a generic LAN with a Unix client, a DOS client, a OS/2 client, Mac client and a Win 3.0 client. Transport protocols are listed (support on top of Netware via NLM's) - AppleTalk, TCP/IP, IPX/SPX, TP4. They also state, "Within the Unix community, NFS has become the industry standard for distributed fie system. Available as an add-on product, NetWare NFS allows NFS clients to view the NetWare file system as an extenion of their native file system, and provides printing support as well as file and resource sharing." COMMENTARY The entire publication is a NetWare-centric presentation of NetWare and LAN computing. It may be the case that additional information is being glossed over (e.g. LAN WorkPlace get three lines of space). However (correct me if I am wrong) it appears that Novell does not offer a way to buy into the "industry stanard for distributed file system... within the Unix community". It is suggesting that the Unix community buy into the Netware "standard". Fine for Netware, mostly unreasonable for existing NFS servers and client applications. It is a step in the right direction, but won't satisfy my needs at a price I am able to pay. Ever onward. barry -- +--------------------------------------------------------------------+ | Barry B. Floyd \\\ barry_floyd@mts.rpi.edu | | Manager Information Systems - HR \\\ usere9w9@rpitsmts | +-Rensselaer Polytechnic Institute--------------------troy, ny 12180-+
phil@brahms.amd.com (Phil Ngai) (03/21/91)
barryf@aix01.aix.rpi.edu (Barry B. Floyd) writes: >The entire publication is a NetWare-centric presentation of NetWare and >LAN computing. It may be the case that additional information is being I've looked at the Novell press releases as carefully as I looked at the Sun PC-NFS press releases and came to the same conclusions you did. Netware is not an ideal solution for us either considering our large investment in Suns. Which is why I am here in this group, trying to get Sun to understand how PC networks work so they can fix Sun PC-NFS. -- The Macintosh makes it easy to do sloppy work.
kdenning@pcserver2.naitc.com (Karl Denninger) (03/21/91)
In article <1991Mar14.051003.6212@amd.com> phil@brahms.amd.com (Phil Ngai) writes: >In article <123@jetson.UUCP> danq@jetson.UUCP (Daniel Quinlan) writes: >|c) Are there any unobvious concerns I should have about adding >|100 pc's to an ethernet network with pcnfs? (I was thinking of putting >|them on a separate subnet to isolate them a bit). > >If you give 100 users a PC-NFS package and one page of instructions, >almost certainly at least one of them will give his PC the same >name as the server and crash the server. If you do all the installs >yourself this is much less likely to happen. (what a bother) > >Stupid users. Bad users. Wish they'd leave my nice Sun alone instead >of causing trouble. They seem to want to get work done instead of >just admiring how wonderful a Sun is. Oh, but there IS a way around this. It's called RARP. I'm sure you have heard of it. We use that here with B&W's product, and haven't had a host name or id conflict yet. Yeah, it means you have to assign the person an IP address or he can't get in at all. It means your ops people have to gen an install disk which does all those things. BUT YOU HAVE TO DO THAT FOR NOVELL TOO! They just call it the "IPX SHELL BUILD". No real difference. B&W's NFS shows up as a "Pc Network" under Windows. It doesn't show drives which aren't mapped. The printer support is only so-so; you can't map new printers, but you can release those you have mapped -- and you do have to use the ".OS2" devices. A real "login" program took me an hour to write for the PC, which handles ALL the specifics (ie: mapping the user's home directory, some shared directories, taking care of mapping default printers, etc). It's nice and automatic, figures out who you are and handles the bloody details (like where your real home directory is, etc). Now, try the following from Novell: 1) Netbios support that doesn't swamp your network with broadcasts. 2) Redirection of COM ports over the LAN to real devices. On B&W's product, you can use it as a "virtual modem" and do "atdt mit-eddie.mit.edu" from your terminal program -- and get connected! 3) Route that terminal session (or nearly anything else) to a machine 1000 miles away -- or anywhere reachable from your network. 4) REAL TCP/IP (or anything else) to a mainframe, without the bull$$$$ that gateways impose on you. (Ever try to transfer 1GB of data from a 3090 through a gateway? It takes a month -- that's only ~50 hours if you can do TCP/IP through a 56k link). 5) Printer service that integrates seamlessly with the rest of your systems, including Unix hosts and Macintoshes (ONE set of printers to buy for everyone instead of one for Unix, one for Novell, and one for the Macs). This alone is such a win in cost that you can spend the winnings ($5k/printer/location not needed) somewhere else -- like on real server horsepower! 6) Real performance from real servers, not glorified PCs running a load they were never intended to handle. Novell can't do any of that (with the possible exception of #6). And there's lots more it can't do either.... and all of it is easy with a TCP product like B&W NFS. Novell will let you see printer queues in action, yes. It will allow you to attach to printers easier, yes. All of which you can provide in about an hour's work with B&W NFS if you care to code a few things up for it. It really is that simple -- or are all the "network managers" out there incapable of using Turbo "C" and doing an hour work of work? The printer deficiencies in seeing the spooler status is more of a LPR/LPD thing than anything else. Have any of you actually looked at that code? It's a gawd-awful mess, frankly, it really stinks. What is needed there is a complete re-write..... which would likely give you a reasonable interface to it so you can do the kinds of things you want. Oh, one more thing. How much memory do you have after you load all of (1-5) above (at least the parts you can get at all), file service, and the like on Novell? Can you get to 590K of free TPA on a '386? Can you have 520K on a 286 (since 286 systems can't do QEMM)? Guess what happens on Novell when you NEED that free low memory. You get to remove IPX from memory, and you have >NO< network. I won't argue that Sun's product was or was not unacceptable to me -- it was. -- Karl Denninger - AC Nielsen, Bannockburn IL (708) 317-3285 kdenning@nis.naitc.com "The most dangerous command on any computer is the carriage return." Disclaimer: The opinions here are solely mine and may or may not reflect those of the company.
cliff@demon.co.uk (Cliff Stanford) (03/25/91)
In article <91074.125204TOMIII@MTUS5.BITNET> TOMIII@MTUS5.BITNET (Thomas Dwyer III) writes: >Ever hear of RARP? We currently have about 300 machines running PC-NFS, >and never had a problem with duplicate host names. The only thing the >machines need to have unique are the Ethernet addresses, which I'm sure is >never a problem with anyone. :-) I've *heard* of RARP but I'd love to have some sort of explanation of how it works. Also, is there an implementation for SCO unix? Regards, Cliff -- Cliff Stanford Email: cliff@demon.co.uk (Work) Demon Systems Limited cms@demon.co.uk (Home) 42 Hendon Lane Phone: 081-349 0063 (Office) London N3 1TT England 0860 375870 (Mobile)