Thanks to all for both the private and public responses to our problem with
an Intermec printer. Maybe I should point out that the printer does not stop
working while spooled output is in progress . It just skips some of the
pages in the middle and then prints some more, loses some, etc.
 
For those who could not read all the responses here is a brief summary.
 
1) It could be a wiring problem.
     We followed the printer recommendations. Only pins 2,3 and 7 are used.
We were planning to test it directly connected to the DTC port to see if
there was any difference but decided to wait for more suggestions first.
 
2) Double check Xon/Xoff and parity settings.
     Done.
 
3) Try a lower speed, the printer might be too slow.
     It is a Bar Code Labels printer. So it IS slow. Even so, if the spooler
and the printer are singing the same tune, the spooler should be able to
recognize when to wait. Since we are using 9600, we may try 4800 but I do
not put much of my hopes here.
 
4) Verify the printer's internal memory. The downloaded label itself could
be filling it.
     Here I must say that the label is very simple. Just  one line of text
and a barcoded number. No fancy fonts, no embeded graphic, nothing worth
considering.
 
5) Change the port to a terminal type, modify the program to include pauses
between groups of pages and optimize the number of seconds of delays by
trial and error.
 
This recommendation comes from someone who has experience with Intermec
printers and working for a company with GOOD technical resources. He says
that the 3000 and the Intermec printers CAN NOT handshake properly.
 
Surprise, surprise. :-(
(I would like to know why)
 
I had the feeling that the real problem was this because we have other Zebra
printers dedicated to bar coded labels and they work fine. I was hoping that
there was a workaround or something overseen, but that does not seem to be
the situation here.
 
We decided then to try a modified approach on the last suggestion for
testing. Since the spooler works fine on a reduced number of  pages
(labels), we can do the testing as it is now with the following changes.
First print the labels to disk. Make a simple program to read the disk file,
find the number of records, and print chunks of 6 pages each with a pause in
between to be adjusted by trial and error. Keeping it spooled may chew up
some resources and we may change it to be a regular terminal port later. At
that time we must reevaluate the delay between print sets.
 
I will post our findings.
 
[log in to unmask]