(Message 139, BBoard-ID: 301) MessageName: (Message sun-spots:139) SUBJECT: PC-NFS REV 3.0 NOT MUCH BETER THAN 2.0 Date: Fri, 15 Jul 88 23:05:47 EDT From: "Greg Hullender" ---------------------------------------- This is an excerpt from a message sent to SUN a few minutes ago. Fourteen months ago, I switched our company from DEC to SUN after ten years as a DEC customer. A large part of the reason for selecting SUN over any other vendor was PC-NFS. Since we've had it, PC-NFS (ALL revisions) has been an unending source of trouble. It seems to be the only product from SUN that doesn't work and that SUN can't fix. We have reported problems with it from day one, and it is not clear that it is better now than when we received it. Next week, most of us will probably be forced to DEINSTALL revision 3.0, because it effectively prevents you from using the PC as a terminal. (See below). Two days ago, I upgraded my PC from PC-NFS revision 2.0 to PC-NFS revision 3.0. At the same time, a few other people at Proximity also did the upgrade, hoping it fixed a few of PC-NFS's problems. We were pleased to discover that two old problems seem to be fixed: first, control-S/control-Q flow control now seem to work. (They corrupted your data before). Also, running the Microsoft C compiler with Telnet installed no longer seems to crash the machine every few compiles. These were welcome changes from before. However, a new bug has been introduced that makes telnet very difficult to use. Every few minutes, telnet stops echoing input for a minute or so. In earlier revisions, this happened a few times a day, but with rev 3.0, this is so frequent that the system becomes almost unusable. (Certainly far worse than before). Anyway, this happens even when I have the 3/280 and the ethernet all to myself, so it has nothing to do with the load. Another serious problem that is still pending is that groups (beyond your main group) are not supported over PC-NFS. We make extensive use of groups in product development, and continue to waste a lot of effort on PC projects working around this defect. In addition, a few annoying minor problems persist, and 3.0 introduced a few new ones. 1) It still isn't possible to mount more than eight file systems, which means we always have to leave some unmounted. 2) Telnet still doesn't let you get at anything like the number of function keys you should be able to. The PC has many keys that do not generate unique sequences under telnet; this makes a lot of things more difficult than they ought to be. 3) Non-DOS filenames under PC-NFS are given random and CHANGING DOS equivalents. We understand why this poses a problem, but the current implementation is much worse than it needs to be. We've offered suggestions before; we can again, if it would help. 4) On PC-AT's and better machines, telnet was sufficiently fast to allow the PC to serve as a person's main terminal. PC-XT's (ordinary 8088-based PC's in general) are FAR slower -- too slow to be used this way except in an emergency. MUCH slower than you'd guess from their relative performance in anything else. Analysis of the ethernet traffic shows significant packet retransmission when the only thing happening is output to a single XT via telnet. New with PC-NFS 3.0 1) The installation program treats "net join" the same as any other net command, which causes it to put such commands into the NFS\NETWORK.BAT file, which is executed BEFORE NFS\DRIVES.BAT. This guarantees that any installation which uses the join command will no longer boot properly, and requires manual editing of two batch files. 2) NFSCONF will not deal with join's in any way, shape or form, which makes it effectively useless for our purposes, even though it is otherwise very nice. 3) NFSRUN fails if it has to read the net command from a RAM disk. Since this speeds up booting, most of our people had installed it this way, but 3.0 won't work that way at all. ==================================================