[comp.os.vms] wollongong tcp/ip vs. CMU/TEK tcp/ip

cetron%ced@CS.UTAH.EDU.UUCP (06/06/87)

At the end of a recent message I made a passing reference to TWG's tcp/ip
package. I have been informed by several people that they considered it a cheap
shot.  If that is the way it was taken, sorry.  I intended it as an upfront
comment that the TWG software is a) expensive and b) not as good as at least
one other package for our distributed environment.

here is the text of a letter I wrote to all of our campus managers after 
evaluating TWG's 2.3 release vs. CMU tek/tcp ip on a head to head basis:


-------------------------------------------------------------
Date: Thu, 19 Feb 87 10:11:28 MST
From: cetron@utah-ced (Ed Cetron)
Message-Id: <8702191711.AA03600@utah-ced.ARPA>
Subject: cmu tek tcp/ip vs. wollongong
Cc: cetron@UTAH-CS


I currently have both the cmu variant of tek tcp/ip and the wollongong tcp/ip 
running on two identical MVII's.  By early next week we (the sys mgrs and the
campus network people) have to decide on one or the other.  If you want to
play with them, please contact me asap.....I have been playing with them for
about 2 weeks and my impressions are as follows:

wollongong:

1. somewhat easy to install, but a few gotcha's  (see 3)
2. uses standard unix host tables.
3. requires a fair bit of configuration
4. will not pass control-c or control-s transparently
5. Is a little slower then the old tek tcp/ip
6. the mailer is a cpu hog - reschedules itself and needs to
	run every 5 minutes to get decent timeliness of e-mail
	and is a terrible kludge.
7. No source
8. costs $$
9. nice manual and some help
10. supports class a,b, and c nets
11. supports shared deqna (i.e. with decnet)
12. supports finger, rwho, ruptime, ftp, and telnet.
13. requires massive/significant sysgen changes to gblpages, irp-,lrpcounts
	and once started, cannot be removed without a complete re-boot and
	must be reconfigured if new layered products are added - this is a
	MAJOR bad point.


CMU Tek tcp/ip
1. VERY easy to install (uses vmsinstal)
2. uses standard tops20/sri-nic host tables (1 table vs unix 3)
3. almost no configuration necessary (three lines of text)
4. will pass control-c and control-s just fine
5. is MUCH faster then old tek tcp/ip and significantly faster then TWG
6. mailer supplied, but interface to others is easy and pmdf will
	be supported since the user interface is standard well
	documented vmsmail.
7. All source supplied (though in bliss)
8. is totally free within the university [note true 'cept for mvII's elsewhere]
9. lots of help, but few manuals
10. supports class a,b, and c with undocumented class d support
11. supports shared deqna (i.e. with decnet)
12. supports finger, ftp, and telnet.
13. requires no sysgen changes to gblpages, irp-,lrpcounts etc.
	and once started, can be removed and restarted without a complete 
	re-boot.



After two weeks, there is no question in my mind that the cmu tcp/ip code is
much superior..... Please send your 'votes' and comments direct to me and
I will collate them with Lee.  

-ed cetron
---------------------------------

as a side light I have the following (new) additional comments:

1. I do thank TWG for giving us a 'loaner' copy of win/vx to test.
2. I was informed by twg, that they do offer substantial discounts to many
	universities. (I have heard rumors of 50%, but that is STILL a lot
	of money)
3. Since many of our sites are smaller machines with fewer users, the slightly
	lighter cpu load of the TWG software is not as important. Several
	other facilities with LARGE sites have said there is some cpu loading
	due to the CMU.

4. Version 3.0 of the TWG software is out and is supposedly better and will
	handle nameservers, the CMU nameserver is in beta test (?) at cmu
	and will hopefully be released soon.

5. My final comment is this:  Several times I have asked the internet at large
	their opinions on TWG software.  I have received many (10's) of very
	negative responses, and roughly 2 positive responses.  One must also
	consider the 'silent majority' concept, but this is a very telling
	statistic.

(also read the figures in a recent posting of comparison between 7 other
tcp/ip's for vms...)

-ed