T.R | Title | User | Personal Name | Date | Lines |
---|
3056.1 | | METSYS::THOMPSON | | Fri Mar 14 1997 10:43 | 11 |
| Robert,
DEC/EDI will try to send up to 64k buffers. i.e. if you have data larger
than that it will go right up to that level.
Perhaps process could not handle it. If necessary we can do cut
down buffers - but it requires a code change.
Do a test with smaller documents and then work up.
Mark
|
3056.2 | no document being sent! | CSC32::R_GOLLEHON | | Fri Mar 14 1997 13:28 | 12 |
| Mark,
At the point this error is occurring, we weren't trying to send a
document. I had entered a bogus trade command which should have failed
with an file not found, but never got that far.
I'll see what OBB had to say and possibly have the customer contact
their TCP/IP software provider and see what they say.
Thanks,
-Robert
|
3056.3 | Escalating | CSC32::R_GOLLEHON | | Fri Mar 14 1997 17:54 | 18 |
| Mark,
I received the following from a contact on the OBB support team. Based
on this, I think I will need to do an escalation.
Thanks,
-Robert
----------------------------------------------------------------------
I took a look at the note file entry. There is a maximum transfer/buffer size
that is dependent on the network transport or operating system used. I believe
that TCPware must have a maximum packet size of 64K. The unfortunate thing
is that it is up to the application to make sure any data it is sending does
not exceed the 64K limit, if it does it must break up the data in to 64K
chunks. In your case, EDI or the EDI application is not doing a good job of
this (and most likely did not realize the it needed to). I am sure that
someone in the notes file will confirm this.
|
3056.4 | | METSYS::THOMPSON | | Fri Mar 14 1997 19:10 | 7 |
|
ahh yes - I just remembered, We always send back 64k regardless
of the actual amount of data.
We need the rules from OBB Engineering.
Mark
|
3056.5 | node name game | CSC32::R_GOLLEHON | | Tue Mar 18 1997 00:30 | 30 |
| I'm not out of stupid questions yet, not by a longshot!
This customer is now trying to switch from IP to DECnet...he can set
the server to use DECnet only, but the client must be able to use both.
We can get the IMF started up on the server with a trade call from the
client, and we connect via DECnet, but for some reason the validation
inside DEC/EDI appears to try to validate the request using the IP
nodename of the client. IE,
VAXB, vaxb.prd.wcus is the client using IP and DECnet
WCEDI1 is the server using DECnet only
I've verified that the proper proxies are in place.
With an entry for VAXB only in the node registry on the server, when
doing a TRADE POST from the client we get a DEC/EDI user not
authorized. If we add an entry for vaxb.prd.wcus, we get a server
error. Looking at the trace, we can see no problems in the log for the
first scenario. For the second scenario, it is trying to use the IP
node name to make a callback to the client and failing for obvious
reasons.
My question is why is EDI trying to use the vaxb.prd.wcus node name for
validation? My guess is that this nodename is passed from the client
to the server, but I don't know where we are getting it from. If so,
how do I get it to use the DECnet node name?
Thanks,
-Robert
|
3056.6 | ObjectBroker gives us the name | METSYS::HELLIAR | http://samedi.reo.dec.com/ | Tue Mar 18 1997 09:02 | 8 |
| Robert,
DEC/EDI obtains the clients hostname via ObjectBroker...ObjectBroker is
supposed to shield us from the underlying transport so we have no idea
whether the name is a DECnet or tcp/ip name, we just use what we are
given.
Graham
|