Will@cup.portal.com (Will E Estes) (10/18/90)
Do TCP/IP and and IBM's LU 6.2 share enough in common as peer-to-peer protocols that it would be possible to build a single API on top of both? I have heard some discussion that IBM's Common Programming Interface - Communications (CPI-C) might evolve into such an approach. Obviously one impediment to a common API is that the two protocols use different naming systems, but assuming you could bridge that difference are the protocols - as protocols only - semantically and syntactically similar enough that one could build a generic model that bridges the two? Thanks, Will Estes (sun!portal!cup.portal.com!Will)
lstowell@pyrnova.pyramid.com (Lon Stowell) (10/19/90)
Yes, you can build a common API. CPI-C on the AS/400 is such a
product, you can run TCP/IP, OSI, SNA from the same API. The
difficulty in naming conventions is hidden in the fine print of
such API's....as are the other differences in these protocol
stacks. All of them provide essentially similar services, but
the LAYER within which a specific service is performed as well
as the implementation method varies widely.
The IBM technique is to note that all of the naming, route
discovery, etc. methods are "implementation specific" and that
they are utility functions performed by the Physical Unit.
Not only do the naming conventions of LU 6.2 differ from TCP/IP,
the techniques for route discovery and routing differ as well.
About the only thing the two have in common is the ability to
provide a high level common API.
Such a "generic" API implementation will never be as efficient
as a protocol specific one, but it sure makes things a LOT
easier on programmers, user's etc. RAM is cheap, and MIPS are
always available. Programmer's blood pressure and sanity are
a little more precious.....
/|
\'o.O'
=(___)=
U
THPTH! ACKHH!jlister@slhisc.uucp (John Lister) (10/23/90)
Interesting thought. The answer seems to be that you could implement something on top of RPC that would have similar functionality to LU6.2 for particular applications. Although IBM has been pushing LU6.2 as a commun- ications panacea, in actuality, most implementations use it for transaction processing, and the orientation is definitely towards Logical Units of Work and the questions of atomicity that Distributed Transaction Processing raise. I have been wondering about starting a project to look at writing an RPC interface for CICS Distributed Transaction Processing for some time but haven't actually done anything about it. Is there anyone out there who has already got their fingers wet? John Lister.