mike lorengo wrote:

> Didier  wrote in message news:<[log in to unmask]>...
>
> >We have pretty big messages to be loaded and we found out that orders
> with
> >a length bigger than 29900 characters do not get read properly by the
> >  java listener program which failed with :
> >
>
> I'm wondering if it has something to do with the limit that can be
> sent in one IPCSEND message, which according to
> http://docs.hp.com/mpeix/onlinedocs/5958-8600/00/00/34-con.html is
> 30000 bytes, you may need to chunk up the sends from the HP? 29900 is
> awful close to 30000...

I do seem to recall that there is a socket limit (not just IPCSEND) of
30000 bytes.  We definitely hit this problem in early testing of our JVM
port.

Now, my recollection may be getting fuzzy here, but I thought we had
worked around the problem within the java.net library code itself -
essentially, the chunking/reassembling of the data into <30K blocks was
done behind your back.  However, it could be that this change didn't get
applied everywhere it was potentially needed.  In particular, we did
something for the Socket class, but it could be that other classes (such
as URLConnection) may have the same problem.  (Then again, they might
all ride on top of the Socket code).

Safest (and quickest) thing is to limit your sends to < 30K, but it may
be something our Java lab (he knows who he is :-) ) might take a look at
   as a potential bugfix for a future release.

Mike

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *