Local printers also created as remote printers on the same machine

[brianjmurrell]

Describe the bug
I have added an HP network printer (port 9100) to cups using hp-setup. That works fine. But I get duplicate printers with my hostname appended that I don't really need or want. It just creates confusion. I.e.

$ lpstat -t
scheduler is running
no system default destination
device for HP_LaserJet_M1522nf_MFP_fax_network: hpfax:/net/HP_LaserJet_M1522nf_MFP?ip=10.75.22.235
device for HP_LaserJet_M1522nf_MFP_network: hp:/net/HP_LaserJet_M1522nf_MFP?ip=10.75.22.235
HP_LaserJet_M1522nf_MFP_fax_network accepting requests since Thu 02 Apr 2026 02:21:47 PM EDT
HP_LaserJet_M1522nf_MFP_network accepting requests since Tue 07 Apr 2026 05:24:11 PM EDT
printer HP_LaserJet_M1522nf_MFP_fax_network is idle.  enabled since Thu 02 Apr 2026 02:21:47 PM EDT
printer HP_LaserJet_M1522nf_MFP_network is idle.  enabled since Tue 07 Apr 2026 05:24:11 PM EDT
$ lpstat -e
HP_LaserJet_M1522nf_MFP_fax_network
HP_LaserJet_M1522nf_MFP_fax_network_pc
HP_LaserJet_M1522nf_MFP_network
Printer_in_Dad_s_Office_using_the_Network_pc

The above *_pc printers are the duplicates being automatically added as "remote" printers.

To Reproduce
Steps to reproduce the behavior:

1. Add a real HP network printer using port 9100
2. Observe that a duplicate print queue is created for it

Expected behavior
No duplicate print queue should be created.

System Information:

• OS and its version: Fedora 43
• Application N/A
• CUPS version 2.4.16

Additional context
I've already disabled cups-browsed to eliminate that as being the culprit.

The duplicate printers are not created if I set Browsing Off in /etc/cups/cupsd.conf but I do want this machine to automatically add print queues for remote printers on other hosts in the network.

$ avahi-browse -rt _ipp._tcp
+ pc_bridge IPv6 HP_LaserJet_M1522nf_MFP_fax_network @ pc      _ipp._tcp            local
+ pc_bridge IPv6 Printer in Dad's Office, using the Network @ pc _ipp._tcp            local
+ pc_bridge IPv6 AirPrint HP_LaserJet_M1522nf_MFP @ pc         _ipp._tcp            local
= pc_bridge IPv6 HP_LaserJet_M1522nf_MFP_fax_network @ pc      _ipp._tcp            local
   hostname = [pc.local]
   address = [10.75.22.1]
   port = [631]
   txt = ["printer-type=0x100E" "Color=T" "mopria-certified=1.3" "URF=V1.4,CP1,W8,PQ4,SRGB24,RS300,FN3" "TLS=1.2" "UUID=4cf6dbcd-83b6-36c6-7ffe-5830dee3176e" "pdl=application/pdf,application/postscript,image/jpeg,image/png,image/pwg-raster,image/urf" "product=(HP Fax2)" "priority=0" "note=" "adminurl=https://pc.local.:631/printers/HP_LaserJet_M1522nf_MFP_fax_network" "ty=HP Fax2 hpcups" "rp=printers/HP_LaserJet_M1522nf_MFP_fax_network" "qtotal=1" "txtvers=1"]
= pc_bridge IPv6 AirPrint HP_LaserJet_M1522nf_MFP @ pc         _ipp._tcp            local
   hostname = [pc.local]
   address = [10.75.22.1]
   port = [631]
   txt = ["pdl=application/octet-stream,application/pdf,application/postscript,application/vnd.cups-raster,image/gif,image/jpeg,image/png,image/tiff,image/urf,text/html,text/plain,application/vnd.adobe-reader-postscript,application/vnd.cups-command" "printer-type=0x829044" "printer-state=3" "product=(GPL Ghostscript)" "note=Hewlett-Packard HP LaserJet M1522nf MFP" "rp=printers/HP_LaserJet_M1522nf_MFP" "URF=none" "Transparent=T" "qtotal=1" "txtvers=1"]
= pc_bridge IPv6 Printer in Dad's Office, using the Network @ pc _ipp._tcp            local
   hostname = [pc.local]
   address = [10.75.22.1]
   port = [631]
   txt = ["printer-type=0x809046" "Copies=T" "mopria-certified=1.3" "URF=V1.4,CP1,W8,PQ4,RS1200,FN3" "TLS=1.2" "UUID=0eadd2c8-926f-3f41-6471-821fb6959254" "pdl=application/pdf,application/postscript,image/jpeg,image/png,image/pwg-raster,image/urf" "product=(Hewlett-Packard HP LaserJet M1522nf MFP)" "priority=0" "note=Dad's Office" "adminurl=https://pc.local.:631/printers/HP_LaserJet_M1522nf_MFP_network" "ty=HP LaserJet M1522 MFP Series Postscript (recommended)" "rp=printers/HP_LaserJet_M1522nf_MFP_network" "qtotal=1" "txtvers=1"]

[zdohnal]

Hi Brian,

is the 10.75.22.1 IP of the machine where you run the avahi command? If not, stop sharing the queues at the machine which is identified by IP, which I guess it is a client who re-advertises the queues.
IIUC you see the duplicate queues on the machine which you want to act as a server - in that case the server does not automatically add the print queues on other hosts in the network, it registers your local print queues which are marked as shared on mDNS, where the other hosts pick them up. So if you stop advertising the queues on clients (which, if I understand you correctly, is unnecessary), you won't get dupes on the server. That can be done by setting that 'Browsing off' on the clients or setting the queues on clients as not-shared.

Please let me know if I'm wrong anywhere - AFAIK that's how Browsing directive should work.

[brianjmurrell]

10.75.22.1 == pc which is the machine that is sharing the printers which are also being duplicated as remote printers with _pc appended. It is not a "print server" but rather just a machine used by a user that also has a printer attached/defined on it.

So I do want 10.75.22.1/pc to both share it's attached printers but also create print queues for any other remote printers on the network.

It just seems to me that CUPS on a given machine should recognize any printers that are being advertised on the network by itself and not add those as remote printers. IOW, surely cups can determine the difference between the printers it is advertising vs. printers being advertised by other machines, yes?

[zdohnal]

Aha, to be honest, it is the first time I see (or I don't recall :) ) that someone wants to use printing environment like this (client-server hybrid) to save time - usually you have server-like setup, where you install print queues permanently and shares them (either by mDNS, or by being them set for sharing and client from outside of local network polls the server for printers), or client-like, where you care only about seeing the printers and print to them.

I have reproduced the behavior - create a perm queue with socket backend, have Browsing to Yes and make cupsd listen to Port 631 instead of localhost, and if avahi is running, lpstat -e shows two queues. Based on the code, we eliminate duplicates only if the queues are both with dnssd backends, so your setup is not covered.

@michaelrsweet is this something we want to support in libcups, or will we stick with clean client/server separation? Such setups will be showing this even with CUPS 3.0, IIUC.

[brianjmurrell]

There's nothing really so special or out of the box about this so-called client-server hybrid configuration.

Imagine a case of a network of workstations where some of them have printers attached. Some people have printers at their desks because they print a lot while others only print occasionally so they can use printers that are attached to others' workstations.

But just because somebody has a printer at their desk doesn't mean that they will never want to print to somebody else's printer. Most people might have black-and-white lasers and a few might have colour printers and so somebody might choose to print to a colour printer somewhere else if they are one that only has a black and white laser.

Just to paint some examples of why this hyrbrid setup is not totally out there.

I am sure there are many other examples of when this hybrid configuration makes sense or is otherwise useful.

[tillkamppeter]

@brianjmurrell in exactly that configuration I also see these unnecessary duplications of entries.

And this configuration is also the most intuitive in a home network. A printer is connected to one computer and then is just shared to the rest of the family.

And I observe the duplication, too ...

There's nothing really so special or out of the box about this so-called client-server hybrid configuration.

Imagine a case of a network of workstations where some of them have printers attached. Some people have printers at their desks because they print a lot while others only print occasionally so they can use printers that are attached to others' workstations.

But just because somebody has a printer at their desk doesn't mean that they will never want to print to somebody else's printer. Most people might have black-and-white lasers and a few might have colour printers and so somebody might choose to print to a colour printer somewhere else if they are one that only has a black and white laser.

Just to paint some examples of why this hyrbrid setup is not totally out there.

I am sure there are many other examples of when this hybrid configuration makes sense or is otherwise useful.

[michaelrsweet]

OK, so first the "lpstat -e" output is listing all the printers that are available, both local and remote. There is no queue associated with the "_pc" printers, so they haven't actually been duplicated (yet).

WRT suppressing these printers on the local machine, there are some things that libcups does to try to do that but we don't suppress all local IPP services because you do want to see local printer applications, Docker instances, etc. I suspect that the URIs that libcups has generated for the discovered shared printers doesn't match the URIs it got from the local cupsd instance.

We might be able to simplify the checks by just looking for a "rp" value in the TXT record that matches on the local printers. This runs the risk of hiding a printer of the same name being served by another local cupsd instance, but you could still manually configure a local queue to point to the other cupsd instance's queue to "fix" that.

[tillkamppeter]

I am in the following situation:

till@till-t14g5:~$ lpstat -v
device for brother: ipps://BRW58CDC9DF8E51.local:443/ipp/print
device for envy: ipps://HP7C5758C0CEC6.local:631/ipp/print
device for KM: /dev/null
device for PDF: cups-pdf:/
till@till-t14g5:~$ lpstat -l -e
brother permanent ipp://localhost/printers/brother ipps://BRW58CDC9DF8E51.local:443/ipp/print
Brother_MFC_L8390CDW_series network none ipps://Brother%20MFC-L8390CDW%20series._ipps._tcp.local/
envy permanent ipp://localhost/printers/envy ipps://HP7C5758C0CEC6.local:631/ipp/print
envy_till_t14g5 network none ipps://envy%20%40%20till-t14g5._ipps._tcp.local/cups
HP_ENVY_6000_series_C0CEC6 network none ipps://HP%20ENVY%206000%20series%20%5BC0CEC6%5D._ipps._tcp.local/
KM permanent ipp://localhost/printers/KM /dev/null
KM_till_t14g5 network none ipps://KM%20%40%20till-t14g5._ipps._tcp.local/cups
lprint_test network none ipps://lprint-test._ipps._tcp.local/
options_test network none ipps://options-test._ipps._tcp.local/
PDF permanent ipp://localhost/printers/PDF cups-pdf:/
testprinter network none ipps://testprinter._ipps._tcp.local/
till@till-t14g5:~$ 

lpstat -v (and also lpstat -l -e) shows that I have 4 permanent queues. By checking /etc/cups/printers.conf I see via the Shared entries that brother and PDF are NOT shared, while envy and KM are shared.

lpstat -l -e (this shows all queues a print dialog is supposed to list, prmanent ones and also IPP destinations for which CUPS creates temporary queues) shows that only for the 2 shared queues I get duplicates: envy_till_t14g5 for envy and KM_till_t14g5 for KM. The unshared queues brother and PDF do not produce duplicates. Also all IPP destinations (Printer Applications, Printers) available appear with only a single entry each, without duplicates.

What should be suppressed is that libcups is listing printers provided by the CUPS daemon it is talking with.

If in an app libcups is pointed to a certain CUPS daemon, XXX, via CUPS_SERVER env variable, command line options, or just the default (XXX can also be just the local cupsd), when it calls the API functions of libcups equivalent to lpstat -l -e, it SHOULD NOT list any IPP destination which is served/shared by the CUPS daemon XXX. So for each destination reported via Avahi the library function first needs to check whether the printer is on the CUPS daemon XXX and in this case skip the printer.

[zdohnal]

There's nothing really so special or out of the box about this so-called client-server hybrid configuration.

@brianjmurrell I'm not commenting whether your setup is incorrect, I was just being honest that I haven't seen that so far or I don't recall seeing it, and I'm not here since yesterday, nor in Fedora or Centos/RHEL. People learn new things every day :) .

Imagine a case of a network of workstations where some of them have printers attached. Some people have printers at their desks because they print a lot while others only print occasionally so they can use printers that are attached to others' workstations.

This is what surprised me - as far as I remember admins use a separate entity as 'print server' (which can be another client, just set to share queues) if they have USB printer or more network printers, where this 'print server' has all printers installed permanently, shares them to other clients and does not have to look up for other printers (because it has them installed permanently, to share them with the rest of room).

And this configuration is also the most intuitive in a home network. A printer is connected to one computer and then is just shared to the rest of the family.

@tillkamppeter but this is not what Brian does. Brian has a printer connected to the machine, and wants it shared, but at the same time look for other printers in the network dynamically. Till, what you described I know. The dynamic discovery requirement what Brian uses at the same time as sharing was surprising to me.

[tillkamppeter]

And this configuration is also the most intuitive in a home network. A printer is connected to one computer and then is just shared to the rest of the family.

@tillkamppeter but this is not what Brian does. Brian has a printer connected to the machine, and wants it shared, but at the same time look for other printers in the network dynamically. Till, what you described I know. The dynamic discovery requirement what Brian uses at the same time as sharing was surprising to me.

Here I only wanted to say that the problem described in this issue already appears in the very simple scenario which many people are using.

[brianjmurrell]

@zdohnal Sorry. I was not trying to imply that you were judging on correct or incorrect configurations. I was merely trying to point out that structured print server / print client environments should not be the only model that CUPS works (without surprise or additional configuring on a per-machine basis to remove surprises) in.

Not every environment has the IT-heavy resources needed to manage such infrastructure. Imagine small professional (i.e. lawyer, doctor dentist, etc.) offices that don't have a dedicated IT budget or IT staff, or even a single dedicated IT person to build and manage such structured environments. They are exactly the kind of environment I described earlier that just scatter work-stations around where they are needed by staff and scatter printers as needed hanging off of those machines and expect to be able to let anyone on any workstation print to any printer depending on their needs and printer availability.

And indeed, to bring this around to @tillkamppeter's example this is how any number of home-user installations work. People at home don't want to have to decide (and then make positive configuration changes to effect) which computers are print servers and which are clients.

For as long as I have been familiar with CUPS it has always demonstrated itself as being a homogeneous system of printing suitable for computers with printers and those without and allowing sharing all of the print resources among them. I feel like this idea of delegating CUPS participants to be servers or clients but not both goes against the grain of CUPS.