home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.os.os2.advocacy:12122 comp.os.os2.networking:2688
- Newsgroups: comp.os.os2.advocacy,comp.os.os2.networking
- Path: sparky!uunet!caen!zaphod.mps.ohio-state.edu!rpi!assela
- From: assela@marcus.its.rpi.edu (A. Andre Asselin)
- Subject: Re: ibm tcp/ip a ripoff?
- Message-ID: <l2-3tbd@rpi.edu>
- Nntp-Posting-Host: marcus.its.rpi.edu
- References: <1iratf$aqq@agate.berkeley.edu> <C0q258.JK5@watserv1.uwaterloo.ca>
- Date: Tue, 12 Jan 1993 16:25:54 GMT
- Lines: 100
-
- eengelke@sail.uwaterloo.ca (Erick Engelke) writes:
- >In article <1iratf$aqq@agate.berkeley.edu> jp1ek@sunc.shef.ac.uk writes:
- >>i am beginning to think ibm tcp/ip is a rip off. my comparative
- >>reference is the wattcp freeware tcp/ip for msdos and tcp/ip for
- >>xenix.
-
- >As the author of the cheaper one you mention, here are some
- >observations and comments.
-
- As a former co-op in IBM's TCP/IP development, here are mine...
-
- >>c) few ethernet cards supported (compared to wattcp with clarkson
- >>packet drivers);
- >Clarkson and later CRYNWR packet drivers have served wonderfully
- >over the years. But probably one of the greatest contributions
- >was Russ's skeleton which allowed anyone to write a decent driver
- >for any new hardware they encounterred.
-
- >To my knowledge, there is no NDIS driver freely available in source
- >code for either DOS or OS/2. From my warped view that makes it seem
- >like every hardware manufacturer has to either contract an overpriced
- >NDIS writing team or learn the hard way.
-
- >And the configuration is pretty gruesome. I find this true of
- >all NDIS installations, who decided we didn't mind rebooting
- >20 times and manually editing scripts just to get the configuration
- >right or if we want to try someone else's TCP.
-
- >I can run DOS based ODI and Packet Drivers fine under DOS VDMs,
- >but my 3c507 NDIS driver kills my modem, go figure!
-
- NDIS is the standard network adapter interface under DOS, OS/2 and now Windows
- NT. There are MANY NDIS drivers available for the various cards, and surely
- more to follow, since this is one area that IBM and Microsoft agree in!
- As far as setup goes, have you ever tried NTS/2 (also called LAPS)? It's
- cake. Click on your adapter, click on your protocol, and bang, you're done.
- Very easy.
-
- >>d) common tcp/ip programmes not or badly implemented (rsh does
- >>not accept piped input or properly return stdout from the
- >>executed command; ftp does not implement piped input/output for
- >>get and put commands; etc).
- >I wouldn't want to brag here because most of the free apps I
- >distribute are pretty haphazzard. My contribution was more
- >a small TCP stack suitable for inclusion into other's programs,
- >like Kermit, WAIS, various 3270 emulation TSRs, several commercial
- >TCP applications, and embedded systems.
-
- >But I too have found the IBM supplied apps very crude and
- >often broken (worst VT100/220 emulator I've seen, TELNETD
- >crashes much more frequently and is less responsive than my own
- >running under DOS!)
-
- I can't vouch for what you're seeing on your machine, but as far as I'm
- concerned, I've never had a problem with any of the Telnet clients getting
- into my campus UNIX computers...
-
- >>i) hard to setup (xenix tcp/ip took me about three hours from
- >>openning the box to a reliable anonymous ftp service; wattcp took
- >>about an hour from loading to basic operation; with os2, i'm
- >>still at it weeks later).
-
- >The reason why WATTCP, NCSA, CUTCP, KERMIT and most other free TCPs
- >are so easily configured is simple - we authors don't want people
- >calling us asking how to get the software working and we don't
- >want to have to document much - hence simple software.
-
- >You might want to note that all the free software use ASCII config
- >files, support some way to do include files so multi-station
- >installs are simple, easily support BOOTP, and seem to need
- >comparatively little documentation to get you rolling. This
- >is not a boast but a reminder that TCP installation does not
- >need to be the pain that it is under IBM's lable.
-
- >Most commercial houses would do well to take a look at getting
- >back to basics and putting the emphasis on the software and
- >ease of installation/use.
-
- I'm afraid that I don't agree with you. ICAT is a very easy to use
- installation tool. If you're setting up your own workstation, you're never
- going to be able to get away from having to know SOMETHING about TCP/IP and
- how your network is configured. ICAT minimizes what you have to know and
- organizes the information well. It's certainly WORLDS better than what I've
- seen on most UNIXs. (BTW, if they can help it, IBM would prefer that you
- didn't have to call them either to install software-- don't forget, every
- one of those calls costs $$$).
-
- >From a programmer's standpoint, I think IBM made a big mistake
- >in their approach to reading/writing data to the tcp connection.
- >Under OS/2 you quickly find that pipes, shared memory,
- >semaphores, queues, everything else maps wonderfully into the
- >filesystem except TCP. Do Novell's, FTP's, Essex's or anyone
- >else's TCPs provide this support? Everything else about OS/2
- >makes BSD code relatively simple to support except TCP, someone
- >really goofed up IMHO.
-
- That's true enough, and I agree with you that it'd be better if sockets
- were integrated into the file system.
-
- - Andre Asselin (assela@rpi.edu)
-