[comp.protocols.iso.dev-environ] Summary of Retix Information Requested

oliveau@oahu.cs.ucla.edu (Greg Oliveau) (09/19/90)

Netters - 

   Thanks for all of your responses to my posting.  A received a lot
more mail than I thought I would, including numerous requests for
a summary, so here it is.

   Retix checks out to be pretty good.  There are a number of other
companies in the same business, but consensus has it that Retix is
about the best you can get.

   I have enclosed my original posting followed by exerpts from 
responses which are separated by double lines (=====).  Text 
in square brackets, [], is my own.  I have not included return mail 
addresses for the responders because a number of people have 
requested annonomy.  If you want to contact the writers of particular 
responders, please contact me by E-mail.  

   Please note that these opinions are those of the responders, not mine,
or my company's unless otherwise noted.  I have no association with any
companies or organizations mentioned herein unless otherwise stated.

   I hope you find this of some use:

> From ucla-cs!oliveau@oahu.cs.ucla.edu Tue Aug 28 17:55:43 PDT 1990
> 
> Retix made a very keen presentation to my company a couple of weeks
> ago.  We are charged with building an ES and an IS and I think
> the Retix source code would help a lot.  Has anyone had experience
> with Retix?  Particular questions that come to mind are
> 
> 	Is the source code readable?
> 	Is it efficient?
> 	Is the 'example code' worth a darn?
> 	How about support from Retix? (it costs a bundle!)
> 	What about overall value?
> 
> Thank you in advance for your help.  *Please* E-mail responses to
> me, then I will post a summary if there is enough interest.  By the
> way, are there other shops in the same business as Retix?  How do
> they stack up?
> 
> Again, thanks!
> 
> Greg.

====================

The 4.3-BSD Reno release contains code for OSI TP4/CLNP, and ES-IS,
and the university of wisconsin is actively constructing an IS-IS daemon.

If you are unwilling to obtain the tape directly from berkeley, any
licensee can give you the OSI-specific code, which is independent
of AT&T licensing.

An interface to ISODE exists, and it has been shown to interoperate
with ultrix/decnet-V FTAM.

====================

ISODE is publically available; you can get it via anonymous ftp

    If you can FTP to the Internet, then use anonymous FTP to uu.psi.com
    [136.161.128.3] to retrieve the file isode-6.tar.Z in BINARY mode from the
    isode/ directory.  This file is the tar image after being run through
    the compress program and is approximately 4.5MB in size.

uunet has made the ATT - free parts of BSD available via anonymous ftp,
they don't have the new release yet, but you could also expect the code
to appear they in maybe about 6 weeks.

====================

OSIware is a possible alternative to RETIX, in Vancouver, BC.  

FON: 604-436-2922; FAX: 604-436-3192.  

There is always ISODE...

====================

We've used Retix on some projects for several years now.  Other companies
to look into include TITN and Touch.  See Data Comm. Buyers Guide or some
other source for addresses and phone numbers.  With some glitches, support
has been good for answering numerous questions we've had, and most of the
software is updated at least once a year (even so, support *is* expensive).
The sample code, like most sample code, is simplistic (e.g., initiator only
or responder only), but good to give you a starting point.  It's useful for
information, it's not enough to base a real system around.  The code is
well-designed for portability, but things are still a lot easier if your
target environment is Unix.  We've ported the lower 4 layers to MS-DOS,
and are working on the upper layers now.  The code is complex but quite
readable, and commented.  The manuals take a black box approach, and don't
explain the inner workings of the layers.  Hooks are in place for things
most people don't use yet, like quality of service, but QOS is not
actually implemented in the source (you could add it).

Several years ago, when we started getting into this, Retix was the only
company that had a reasonably mature full stack.  I can't compare value
at present.  Similarly, can't help on efficiency.

====================

Greg, I really don't know, since I haven't looked recently, and when we
first started with OSI they were about the only game in town.  At that
time, years ago, Touch products were still about Alpha test.  Now that our
programmers and software people know Retix software, we're not looking to
change.  One thing for sure, if you're porting source code over, it's
a considerable effort, not just load and compile!  Still far better than
trying to re-invent the stuff!

====================

I haven't dealt with them personally, but Encore Computer Corp. has
just released a multithreaded GOSIP implementation that is based on
Retix's products. They (Retix) seems to specialize in OSI software
and the products looks solid. I haven't heard any complaints about them.

>By the way, are there other shops in the same business as Retix? How do
>they stack up?

I believe that Lachman (now Interactive) has made some attempts at
producing an OSI protocol stack. I don't know much about it, though.

	The code was quite readable when we had it here...  clean state
machine, short functions, block comments per function.  Too bad we couldnt
use it (politics basicly).

====================

My ISODE experience is rather indirect, and largely colored by being
Marshall's mentor since 1982, so I might be biased.  yes, it is entirely
Upper Layers at this point, with release of BSD 4.4, it will include the
Lower Layers too (the LL code comes from Wisconsin, courtesy of IBM).
But, BSD 4.4 ain't out yet!  Don't know when it will be out, either.  

I guess I misread what you were asking for, and did not perceive the LL
orientation.  I expect that RETIX is your best bet, but I know it is
indeed very expensive.  

====================

>>>>	Is the source code readable?
Yes, is it. You must hold the technical documentation about, anyway.
The right problem is TO HAVE this documentation, expecially if 
you have to write deeply on it.
>>>>	Is the 'example code' worth a darn?
It seems the example code is just an example, oops! just a very 
poor example.
>>>>	How about support from Retix? (it costs a bundle!)
I have not a good experience with this support. My contact is 
Retix Ireland, US support team is different,maybe.

====================

I visited retix a while ago, and looked at their code.  I think this is
worth doing.  I thought it was OK to look at and reflected good 
programming practice.  I looked at the transport code only, though,
so again, you might want to do this yourself (if they let you).

> 	Is it efficient?

I haven't run it.  

> 	Is the 'example code' worth a darn?

The example code which provides the MAP API for FTAM didn't have
asynchronous operations of the nature implied by the MAP API committee
(i.e., async means "start a background operation" rather than the
unix NDELAY or NONBLOCK semantics).  That may have changed.

> 	How about support from Retix? (it costs a bundle!)
> 	What about overall value?

I think, from all I have read of other [..] people's opinions, that the 
retix code provides a good start.  You may want to make modifications
to it based on specific requirements you have. 

> 
> Thank you in advance for your help.  *Please* E-mail responses to
> me, then I will post a summary if there is enough interest.  By the
> way, are there other shops in the same business as Retix?  How do

Yes, there are.  One of them is Marben, which is located in Paris
(France).  Another is Spider (Spyder?) systems, which I think is
in the UK (but it might be just the east coast of the US?).  Another
shop I have heard of is LDR (LDR systems?) in the UK.  There was
another one I heard of located somewhere in the southwest (Arizona?)
I think it was sawtooth systems or something like that.  But it's
been a long time (5 years) since I saw anything of those
last two.  (I haven't been looking, either.)  My shot at things is
that Marben is excellent as is LDR.  I don't know enough of the
others to comment.

====================

        Hmmm...  well I didnt spend a lot of time working with the code
personally, but the listings I looked at looked more comprehensible than a
lot of code I've seen for the reasons I mentioned (plus consistent coding
style).  The guy who ported it to our machine and worked on STREAMS-izing
it (except the ethernet interface) didnt complain...  and some other Unisys
types in Utah, who work with our present product (from Marben) have seen
the Retix code and seemed impressed.

        However, times change and code can get messed up.  As far as I
know, the main competition is Marben and Touch.  There is also the PD UWisc
code, which is supposed to go into 4.4BSD.

====================

I think the code is ok, about the philosophy, and it follow a right scheme.
The reader should be an experienced programmer, a C-begginner can have,
maybe, a little shock. The support for us is not fast, maybe because 
Retix Ireland is not a development place, or maybe in Ireland the 
language is not a BBC language like I speak (??????!!!!!!?????????).

I'm in a project team working on FTAM and VT, so I can not say you a lot of
things about yours interests.

The only things that I can say about Retix is that is a leader in this field,
at the moment I'm a consultant in [..] and retix is very present here.
So, it seems everythings goes well. It should be better get a comparison
with some other store. Anyway I think is the only one with a large listing
of products.
Finally, I think Retix will be present in the [..] future.

The other store I know is located in Italy and the name is TECSIEL.

====================

Network Mangement code.  I personally don't have first hand experience with
the code, but I know that when we bought Nodal Management code, we  were 
allowed to read it for supportability and such.  I think it's important for
you to be able to take a look at it and analyze the areas of expandability,
supportability, and how readable it is. 

>	Is the source code readable?
>	Is it efficient?
>	Is the 'example code' worth a darn?

  By reading the example code, you can probably get a good ideas about the 
  two questions above it.

  I beleive 'Touch' in Santa Cruz sells their code also. It is mainly based
  on 'Marben' code who is in Grenoble France. I imagine it is very good code 
  considering the fact that Europe is about five years ahead of the U.S. in
  networking.

====================

   I don't know much about the Retix package for Es / Is. 
   Are you aware of the OSI offering of Touch Inc / Marben ? They have
   a great Es / Is package (among many others).

====================

. . . I did an evaluation
of OSI protocol suppliers recently.  I found that Retix appeared to be one of
the two best options.  It depends a bit on what you are building, and which
bits you think you would best do yourself.  At the lower layers, especially
on LANs for DOS and OS/2 workstations Retix new LAN transports look second
to none.
At the upper layers, Retix have the broadest set of shrink wrapped, and source
code for applications.  

The other vendor I like is Data Connection in the UK.  Though relatively
unknown, they are pretty solid.  They have a 1988 stack and 1988 X.400 MTA.
Also some useful tools for building OSI applications like an interesting
ASN.1 server, and an object oriented database server.  Edinburgh Uni computer
centre are using Data Connection stuff for some workstation based OSI App.
development they are doing for the UK Janet.

They have portations to many platforms, (Unix, OS/2, DOS, Windows, other)
but don't sell shrink wrapped in the end user market.

Other people we looked at are:
TITN - San Franscisco area (popular X.25 provider, own lower layers up to 
session, agent for Logica X.400 s/w)
Touch/Marben - San Fransisco area too,
Spider - Edinburgh  - (up to Transport layer only).
OSIWare - Vancover - 1984 X.400, and working on X.500

Others I got info on but didn't look closely at was Net-Tel in the UK.


====================

We have no direct experience with Retix source code but we let the Sema
group port the Retix MHS software on Sun X.25. This was about 2 1/2 years
ago. The quality of the software was far less then optimal by then. We
found a number of bugs in MHS and session. But they are almost the only
ones on the market and they have been improving their stuff for some years
now. I also heard from other colleagues that the software was not perfect.
Given the amount of 'B-test' sides Retix has it is probably the best you
can get. The support of Retix was (as far as we perceived) reasonably well.

The only other supplier is OSIWARE I think in Canada. They survived from an
earlier company that was businesswise less succesful.

====================