WANCHO@WSMR-SIMTEL20.ARMY.MIL ("Frank J. Wancho") (05/17/91)
I've reviewed the archives, looking for observations on the KNET software, and found mixed reactions, with some of the comments from several years ago quite negative. What's the current status? Has it been brought up-to-speed, or should we still look elsewhere (given that we were given the hardware)? --Frank
c_bstratton@HNS.COM (Bob Stratton) (05/18/91)
Date: Thu, 16 May 1991 22:38 MDT From: "Frank J. Wancho" <WANCHO@WSMR-SIMTEL20.ARMY.MIL> I've reviewed the archives, looking for observations on the KNET software, and found mixed reactions, with some of the comments from several years ago quite negative. What's the current status? Has it been brought up-to-speed, or should we still look elsewhere (given that we were given the hardware)? I had client last year that was in the throes of bringing up KNET with both MVS and VM machines. As I understood it, Spartacus was rewriting code specifically for this client. I witnessed around 8 revisions in 3Q and 4Q of last year, but their stuff still didn't resolve to DNS servers. I worked in detail with the SMTP implementation, and I wasn't impressed - I'd look at the U. Wisconsin TCP/IP implementation. Bob Stratton | Stratton Systems Design| SMTP: strat@gnu.ai.mit.edu, c_bstratton@hns.com Alexandria, Virginia | PSTN: +1 301 409 2703 "Personally, I think the DNS administrative interface was designed by the IRS." --Mark Beyer
NJG@CORNELLC.CIT.CORNELL.EDU ("Nick Gimbrone ", WIRDI: Whatever Is Right, Do It) (05/31/91)
On Fri, 17 May 91 14:07:36 EDT you said: > Date: Thu, 16 May 1991 22:38 MDT > From: "Frank J. Wancho" <WANCHO@WSMR-SIMTEL20.ARMY.MIL> > I've reviewed the archives, looking for observations on the KNET > software, and found mixed reactions, with some of the comments from > several years ago quite negative. > What's the current status? Has it been brought up-to-speed, or should >servers. I worked in detail with the SMTP implementation, and I wasn't >impressed - I'd look at the U. Wisconsin TCP/IP implementation. I was recently speaking w/ someone who have both the KNET and IBM TCP/IP products. IBM's product is a fixed up version of the Wisconsin work (function added, performance improved, etc). As of early this year not only does the IBM implementation far surpass KNet in function, it is also now running faster than KNet. I believe that most IBM VM and MVS sites feel that there is but one viable alternative today - IBM's TCP/IP for VM and MVS.
gwilliam@SH.CS.NET (George Williams) (06/08/91)
Date: Thu, 16 May 1991 22:38 MDT From: "Frank J. Wancho" <WANCHO@wsmr-simtel20.army.mil> Subject: KNET 200 I've reviewed the archives, looking for observations on the KNET software, and found mixed reactions, with some of the comments from several years ago quite negative. What's the current status? Has it been brought up-to-speed, or should we still look elsewhere (given that we were given the hardware)? --Frank [ Views and Opinions Expressed are mine. ] [ This is neither endorsement, critique, or "expert" ] [ analysis - just empirical fact. ] The box is old and the software dated, but you can use the K200 as a local ethernet channel attachment running TCP/IP. Please note:( from prior efforts in this area) () I would not use the version you have. () It has proven inter-operability, and is a viable alternative, if you have support or source code ! () If you don't have to program it and are just talking using FTP and Telnet ( some of the timers and algorithms we found to be rigid, but that was back in '89-90 ) then you can iteroperate with quite a number of vendors. () The programmatic interfaces required much debugging. I built on top of these the second time ( implementation ) around. Below are some remnants of mental notes regarding the above. The version you mentioned , 2.7 (software), is outdated and no longer supported ( Unless you bought the source ). It had quite a few bugs too.This was true as of early '90 time frame. I am told 3.0 is ready with 3.1 in beta by people doing work in this area. There were a great many functional enhancements we proposed ( at the time ) that went into 3.0 and subsequent point releases.There was a major architectural change that allowed one KNET virtual machine to have multiple ethernet drivers ( node names and corresponding IP addresses ). Thus, you could have an IBM host addressable by more than one IP number. This becomes a requirement on large mainframes ( e.g.3090's running . It was proven that the K200 and not the IBM channel would be the bottleneck , when it came to service access. Application servers running on top of KNET eventually needed more than one K200 as the number of clients (seats) you had doing large, megabyte(s), data transfers increased. Configuration was not a problem as long as you had a SUN workstation that did routing on your network. You just pointed your KNET virtual machine ( "vianode" ) at it, for routing purposes. IBM to IBM routing was taken care of by KNET. They had RIP, but we never configured or used it. Mail worked via the 'Notes' facility and showed up as a file in your virtual rdr. You had to have several SMTP VM's running. The programmming interfaces were written in assembler on the version you have. Version 3.0 had support for 'C', I think. They were also event driven with no socket calls. Process to process event notification was via ECB posting and IUCV calls. BTW, the SAS compiler C level has support for both of thes functions. We went with the low-level calls because of the need for multiple sessions from a single server. And the interface portions of the servers had to port to multiple systems. KNET TCP/IP was innovative ( circa '84 ) as a result you might find it a bit rigid in it's implementation. IBM's implementation is working to be RFC compliant and that work appears to be ongoing. It's not TCP/IP heaven ( there are bugs from what I here ). But no doubt, the support and service will be there. George Williams