Very Computer

This article is about jobs remaining in the queue after being
printed.  The bug was reported and the call was escalated to
engineering.  Engineering asked for a justification and then
told me that they would not fix the bug.

(in comp.unix.solaris)

The way Sun engineering has handled printing bugs has been terrible:

1.  Why does engineering care about business justifications?  It seems
    to me like engineering should be solving technical problems, they
    are not in sales.  Ah, maybe that is the problem, we have a bunch
    of salesmen working on Sol 2.1 printing.  That would explain the
    quality of the printing software and their ability to fix the bugs.

2. How can engineering have the nerve to arbitrarily decide that they
   are not going to fix the junk that they put out in the first place.  

3. When is engineering going to start distributing patches for the
   Sol2.1 printing bugs?

4. Are any of the bugs going to be fixed in Sol2.2, or should I start
   writing my justification letter for printing bugs in Sol2.2.

5. Customer support refused to give me the name or number of someone
   in engineering.  Customer support is going to interact with
   the customers, but they cannot answer my questions.  (by now,
   they probably don't want to talk to me anyway).

Sun forced the customers purchasing Classics and LXs to run Sol2.1.
Our Classics and LXs sat in boxes for 3 months, we decided not to
give the machines to the users because there were too many printing
problems.  After Sun didn't solve one serious bug, we identified the
problem ourselves, we modified the source code and worked around it.
Today, printing still hangs, and the machines have to be rebooted for
printing to resume (description below).  It would be so nice if Sun
forced us to run an OS that handled printing properly.

We are wondering if Sun figured that no one was going to print from
Solaris 2.1.   After all, Newsprint wasn't supported, nor was the
parallel port, or any of the multiport serial boards.

Description of lpNet problems:
Several people have mentioned that there are endless connect/disconnect
messages in in /usr/spool/lp/logs/lpNet.  The log files grow, but
printing still works.  We have a different problem, lpNet tries
to reconnect, but it never succeeds.   Printing stops working.
 We see these msgs over and over:
      lpd retrying connection to (print-server-name)
You can do an lpshut, kill any lpNets that are alive, and restart
lpsched.  But this does not fix the problem, the new lpNets continue
to try to reconnect.  The machine has to be rebooted for printing
to work.

Quote:

> Alternative: Buy some cheapo 486s and run LINUX.  OK, maybe not as complete
> as Solaris, but you get the SOURCES to fix the problems YOURSELF!

We have the Solaris source code and we have spent too many hours
tracking  and fixing the problems.  This should not be the customer's
responsibility, Sun should be fixing the bugs.

Edgar

From what I have read in the bug report, it appears that you were
misinformed.  It appears to me that the situation is that Engineering
will not fix the bug *in*the*release*about*to*come*out*.  There is a good
reason for this.  If you put a fix into the system at the last minute
to fix something that is an annoyance (i.e., it has a workaround), you
are taking a risk that you will introduce a problem which is a serious
impediment to using the system (i.e., you can't use the system at all,
and there is no workaround).  Last minute fixes must have their risk
versus benefit weighed.  This is where the business justification comes
in.  The business justification looks at how large is the potential
benefit.  You compare that with an engineering analysis of the risk of
the problem considering how easy it would be to break something else and
how thoroughly you believe you can test the relevant component and those
with which it interacts in the time you have left before the release.  In
this case, it appears that the fix was not considered to be acceptable
for this release and has been scheduled for the *next* release.  Note:
I can't promise it will make the next release either, but as somebody
watching this from far away (from the Solaris X86 area), it appears to
me that that is what will happen.

Quote:>   Customer support refused to give me the name or number of someone
>   in engineering.  Customer support is going to interact with
>   the customers, but they cannot answer my questions.  (by now,
>   they probably don't want to talk to me anyway).

What happens if customer support starts handing out phone numbers to
everyone?  Engineering would soon be too busy answering phones to fix
bugs.  I'm glad to have some peace and quiet.  I'm taking some risk here
letting my e-mail address out because I really want to help anyone I can.
But I hope I don't start getting deluged with requests because I can't
help everyone.  I have almost no power to do anything in the Sparc world.
In the X86 world, I can try to connect you with appropriate people.  But
I certainly don't promise that I can help everyone who writes.  If you
want to make sure that you have somebody properly tracking your requests,
that's what the customer support people are there for.  I would rather
answer a question about a printing bug once to a customer support person
(plus once here on the net) than having to answer the same question for
each and every customer who runs into the problem and happens to have my
number from some past problem.

     Richard M. Mathews                 Lietuva laisva = Free Lithuania
                                        Brivu Latviju  = Free Latvia

                                                WE  DID  IT!!!

First, I would like to explain that I am not only complaining because
Sun did not fix one small bug.  I am complaining because we have hit
at least nine bugs, and until I posted this article, I did not even
get one useful answer from Sun (let alone patches/fixes). The only
answer that I got was "we are pushing your call back, we won't work
on your bug".  The bugs are listed below and I had already posted them
before.

Second, this is not a case of adding a fix "at the last minute".  Some of
these bugs were reported two months ago, and there still are no patches.

Third, this bug may sound like an "annoyance" to you, but this bug
combined with the printing hangs is a problem.  I never know if the
jobs made it to the print server or if they remained on the client.  
I am also not the only one on the net complaining about it.

The good news is that I have been contacted by Sun engineering and that
I finally got some answers from the people who are working on the
printing software.

Yes, it looks like customer support misinformed me.  The bug about
jobs remaining in the queue (bugid 1075755) will be fixed for Sol2.1
in patch 100863 (the version # is probably 4).  this patch should
hopefully be available in a few weeks.

lpNet connects/disconnects (bug id 1094797) over and over, and
the log files grow.  this bug is fixed in patch 100863 version 3.
this patch should be available very soon.

Quote:

> >   Customer support refused to give me the name or number of someone
> >   in engineering.  Customer support is going to interact with

> What happens if customer support starts handing out phone numbers to
> everyone?  Engineering would soon be too busy answering phones to fix
> bugs.  I'm glad to have some peace and quiet.  I'm taking some risk here

Engineering is now telling me that I was asked for a business justification
because of "a miscommunication between support and engineering".  I agree
that engineers should not be answering phones.  

But it has not been easy to deal with customer support.  I have reported
about 10 printing bugs to Sun,  I am talking to 10 different people.  
Customer support reps will only deal with one problem at a time.  If you
report a bug, and if it could be a combination of several issues, they will
split the problem and ask you to open a new call and have someone else
look at the related issue.  When I could not get any help from customer
support, I called the "Solaris Support Center", OPCOM, (1-800-363-6200).
I spent about 20 minutes explaining a bug to a tech support person.  At
the end of the conversation, the person told me: "Oh, by the way, tomorrow
is my last day" ...

If customer support cannot resolve my problems, and if they are misinforming
me, there should be a way of reaching someone who knows what they are
talking about.  (besides posting rude articles on the net).

Edgar

DESCRIPTION OF BUGS:

SEVERITY: high
1.  lpNet cannot reconnect to the print server, printing hangs.
    /usr/spool/lp/logs/lpNet shows:
      lpd retrying connection to (print-server-name)
    the machine has to be rebooted for printing to resume.
    reason: lpNet is trying to bind to a reserved port, but no
    ports are available.

2.  on diskless clients, lpNet cannot read files in /usr/spool/lp/temp.  
    /usr/spool/lp/logs/lpNet shows:
      spool file read error (14)
    printing hangs.

SEVERITY: medium or low
3.  lpNet connects/disconnects from the print server.
    /usr/spool/lp/logs/lpNet grows constantly.

4.  When you spool a file to a bsd print server, lpstat -t or lpq do
    not show you the name of the file that is being printed.

5.  lpstat -t or lpq on a print client do not show the files in the
    queue of the print server.

6.  the bsd commands (lpr, lpq) use "lp" as the default printer.
    instead of using /etc/lp/default.

7.  the cancel command does not remove jobs from the queue.

8.  fast filters on a Sol2.1 client should run on the client and send
    the output to a  BSD server.  But the filter does not run locally,
    it just passes the filter information to the bsd machine, and the
    bsd machine does not know what to do with it.
    Example:
    we have a troff file and would like to use the dpost filter which
    comes with the system: lp -T troff  filename
    The troff file is sent to the bsd server without being processed.
    This is an example in "SunOS 5.1 Setting up user accounts, printers
    and mail", page 418.

9.  jobs not removed from the spool directory after the files are
    printed.

SO WHAT IS THIS WORKAROUND ?

At the moment Solaris 2.1 can not reliably print to BSD style LPDs.
This means the entire installed base of existing SUN Unix machines,
as well as most standalone printer Servers, Novell boxes, PCs acting
as printservers (PC-NFS LPD), OS/2 systems (FTP or IBM TCP/IP stacks)
and a wide range of other systems that recognised the BSD derived LPD
system as the de-facto standard for network printing (Including IBM
HP and SCO).  This is about as big an annyance as finding your Car
no longer has wheels.  We Sell PC-NFS for you and Server configuration
is one of our biggest headaches. The fact that Solaris shipped w/o
a PCNFSD and that SunUK continued to deny the existence of this after
we sent them a copy placed on bcm.tmc.edu by Mr. G. Arnold.

Quote:>impediment to using the system (i.e., you can't use the system at all,
>and there is no workaround).  Last minute fixes must have their risk

If you can't print order, invoces and other forms you can't use the
system.

Quote:>>   Customer support refused to give me the name or number of someone
>>   in engineering.  Customer support is going to interact with
>>   the customers, but they cannot answer my questions.  (by now,
>>   they probably don't want to talk to me anyway).
>What happens if customer support starts handing out phone numbers to
>everyone?  Engineering would soon be too busy answering phones to fix
>bugs.  I'm glad to have some peace and quiet.  I'm taking some risk here
>letting my e-mail address out because I really want to help anyone I can.

I know that it is not your job to answer these questions BUT Sun
should make sure that the people in Customer Services are able to
and can get things sorted out. If this means that Engineers have to
spend a 3 month detachment every three years or so making sure that
Customer Services know how, Then it will also mean that the Engineers
will start to know what is important to the real user base as well.

Thank you for at least sticking your head up.

(Our problem is that the idea of a workgroup of PC's round a SUN has been
undermined by Solaris)

Would someone please refresh my memory about this printing problem?
I've been successfully printing from Solaris 2.[01] <--> Solaris 1.x for
quite some time.  Is there something I'm doing right?  Or am I just
using up some good luck?

---
 Richard Elling, Manager of Network Support, Engineering Administration

 Show me a one-word commercial and I'll show you an adverb.

Install one of the following patches:
Solaris 2.1: 100863-04
Solaris 2.2: 101025-01

Engineering never refused to work on a fix.

Engineering just needs to know the impact of problems so it
can prioritize which things to work on first.   They cannot
work on everything all at the same time. The passage
of such information that can be used to prioritize is called
"justification".    Customers are not asked to supply this, and
if one was, it was a mistake which has been pointed out.
--------------------------------+---------------------------------------
Mark Barnes, System Engineer    |  <insert standard disclaimers here>
SunSoft                         |
Corporate Technical Escalations |  I speak for myself, an individual,
Menlo Park, CA, USA             |  not for the company for which I work.

--------------------------------+---------------------------------------

Quote:> Install one of the following patches:
> Solaris 2.1: 100863-04
> Solaris 2.2: 101025-01

It is not easy to get these patches.

100863-04 solves several of the printing problems in Sol2.1.
But this patch is not official yet, and it is not in the OSS
database (the Sunsolve online database).  100863-03 is not
available either.  If you call the 800-USA-4SUN# and ask for 863-04,
you will get 863-02 which does not help.  If you need 863-04, please
open a call with Sun.  It really seems to take a long time for a
patch to become official.  On the other hand, I have been getting
a very good response from an engineer in the Solaris printing group.

We upgraded to Sol2.2.  The printing problems of Sol2.1 are still
there, so a patch needs to be applied.  863-04 cannot be easily
applied since it is a Sol2.1 patch.  I tried to get 101025-01 from
Sun, and this patch is being held up, because there were problems
during testing/evaluation.

Edgar