T.R | Title | User | Personal Name | Date | Lines |
---|
3974.1 | | RMULAC::S_WATTUM | Scott Wattum - FTAM/VT/OSAK Engineering | Tue May 27 1997 09:51 | 6 |
| Unless the application used undocumented and unsupported QIO calls into DECnet
(such as some of the special NETACP functions) the goal of DECnet-Plus was to
provide a fully backward compatible interface. So in general, the answer would
be yes, all they should need to do is re-compile and run.
--Scott
|
3974.2 | | BACHUS::KNEUTS | Erwin Kneuts DTN 856-8726 | Tue May 27 1997 12:20 | 7 |
| Previous entry (3973 hanging NSP ports in OSI 7.0) is an illustration of the
fact that migration from phase IV to Decnet + is not as smooth as we would like.
It is exactly what we did, recompiling an application using $QIO interface under
5.5-2 and decnet IV to 7.0 and decnet+.
Still looking for some answers there, so any suggestions are welcome.
Erwin
|
3974.3 | Why did you recompile? | TWICK::PETTENGILL | mulp | Thu May 29 1997 00:19 | 20 |
| The applications should run as is. The compatibility is at the binary level.
What has seemed to be a problem is that certain fields in the NCB that should
have been null were filled in with something that DECnet-OSI considered
garbage, but that DECnet-Classic ignored. If you look thru this conference
you will find reports of such problems and it seems to me that all were
reported as fixed in a future release.
If you're having problems, its probably necessary to find the offending
service call and make it conform to the documented interface.
Another source of problems might be making a $QIO call and then not $WAITing
for it to complete. Somethings that were simple and could be handled
synchronously became complex and so they were really asynchronous and you
really have to wait for them to complete before using the results.
Again, it comes down to implementing to the documented service call.
The reason to use the $IPC interface is that it offers more options and
it was designed to be more efficient for the network services then QIO.
|
3974.4 | And In One QIO case an IOSB is Mandatory, else System Crash. | EICSMS::CELLAM | | Mon Jun 02 1997 05:01 | 9 |
| On the case of QIO's we have found a case where the Iosb field is
no longer optional, you have to give a valid address, if not then the
system crashes. I can't remember where DECosap engineering said this
happened, it was definitely in OSI Transport and I thing it was when
breaking a connection. This problem only started occuring with DECnet
Plus 7.0 and is still there in 7.1. I'm not sure why the DECosap
engineers didn't inform the DECnet Plus engineers.
Chris
|