Roadmap

What the library does today, what it does not, and which of the gaps are meant to close.

Today

Area	State
Printing	Raw payloads with a content type, IPP and raw TCP transports, the operating system spooler on Windows and on Linux and macOS, a CUPS server over the network from any operating system, job queues and job progress.
Printer discovery	mDNS/DNS-SD browse, which needs no host list; a TCP probe of explicit hosts; the operating system spooler.
Printer manager	One entry point that runs every source, groups the channels into one device per physical printer, and prints, reads a status or watches a job by identifier.
Printer capabilities	Media sizes, trays, media types, output bins, qualities, pages per sheet, and the defaults the printer applies when a job asks for nothing.
Printer status	Over IPP, and over SNMP version 2c, which adds the serial number and the page count.
Document formats	ZPL, EPL, CPCL and ESC/POS mapped to a document-format the peer accepts without converting the job. A PDF prints five ways; see Document formats.
Page placement	The five PWG fit modes, an anchor and an offset in a physical unit, a smoothing switch, a media size the printer has no name for, and a media size taken from the document. See Page placement.
Raster	PwgRasterWriter, PngWriter and RasterCanvas are all public, so an application with its own rasterizer can use them.
Diagnostics	Each state reason separately, the readable messages, the transport that answered, structured failures, an optional capture of the raw IPP answer, and a graded log over the whole stack.
Protected documents	A job carries the password that opens a PDF. It reaches the renderer and goes no further: no protocol carries it and nothing logs it.
Scanners, other peripherals	Not implemented.

Planned

Scanners — start a scan, retrieve the image — and further peripherals, behind the same shared device interfaces the printing code already uses.

Three smaller gaps on the printing side:

• An LPD transport. _printer._tcp answers are discovered and then discarded today, because no transport here writes the LPD message format, so a printer that advertises only LPD is out of reach.
• SNMP version 3, which adds authentication and privacy. Only 2c is supported.
• IPP notifications (Create-Printer-Subscriptions, Get-Notifications). Job progress is read by polling today, which works with every printer and needs no notification channel.

USB printers

This is not a gap, and it is not planned. The library does not talk to a USB printer directly: there is no USB transport, no USB endpoint and no usb scheme. Print to one through the operating system queue, with a spooler:// identifier.

The reason is that each operating system already claims the device with its own driver, and taking it away breaks the print path the machine uses for everything else:

System	What owns the device	What direct access costs
Windows	usbprint.sys	libusb needs that driver replaced by WinUSB or libusbK, which breaks normal printing for every other application.
Linux	the usblp module, as /dev/usb/lp0	libusb must detach the kernel driver, which removes the device node and makes CUPS fail. It also needs udev rules.
macOS	the CUPS usb backend	There is no device node. Direct access needs IOKit interop and takes the device away from CUPS.

Grouping still works across the boundary: a CUPS queue reports device-uri as usb://Zebra/ZTC%20ZD421?serial=X4TY012345, and the serial number read out of it becomes the same device key an IPP or SNMP read produces. See Spooler and CUPS.

What has met real hardware

The IPP and CUPS channels are tested against real servers on every build: the integration suite runs a CUPS daemon and ippeveprinter, the CUPS project's own IPP Everywhere server, in containers, and prints to them.

The Windows spooler has printed on real Windows hardware. Everything around the native calls is tested on Linux through seams, so the buffer protocol, the page loop and the error paths run in the ordinary suite.

The page placement work has not met hardware. It is arithmetic with unit tests, and where a page actually lands is a question only a ruler answers. Measure a test print before relying on an offset in production.