How a Repo Name Stalled Linux Drawing-Tablet Drivers

There's a particular kind of bug that no compiler will catch: the one baked into a directory name a decade ago. Linux's drawing-tablet support has one of those, and it just cost the open-source community a real shot at first-party vendor collaboration.

The short version, drawn from artist and reviewer David Revoy's account of his back-and-forth with tablet makers: after years of volunteer work dumping hardware specs and feeding them to kernel developers, he finally reached a technical contact who turned out to manage drivers for four brands at once — Gaomon, XP-Pen, Huion and Ugee, all sitting under Shenzhen Huion Trend Technology. The unification itself wasn't a surprise; their proprietary Debian packages had long looked suspiciously alike. Then came the rejection. Not over money, not over engineering effort. The deal-breaker was that the open-source infrastructure they'd be contributing to is branded after Wacom, their largest competitor.

My take: this is a governance and branding failure dressed up as a business decision, and it's entirely fixable — but only if someone with budget decides it's worth fixing. The technology is the easy part. It already works.

The excuse doesn't survive a hid-recorder dump

The rejection email leaned on two reasons. One: limited upside, because even after adding support, the device would still surface through Wacom-named plumbing. Two — the one that gets repeated — sharing device specs "directly with Wacom" is off the table.

That second reason is, frankly, theatre. Modern graphics tablets are USB HID devices, and their report descriptors are not secrets. Anyone running Linux can point hid-recorder at a tablet and dump everything the driver authors need — the same trick Revoy himself uses on every review unit. If an artist with a borrowed tablet can pull the specs in an afternoon, so can any competitor's engineer. There is no NDA-grade IP being protected here. The specs are already in the open the moment the hardware ships.

Which means the first reason is the real one. Brands don't want their name committed into repositories stamped with a rival's logo. It's an optics problem, and optics problems are unusually durable because no engineer can patch away an executive's discomfort.

Why "wacom" is stamped across the whole stack

Here's the irony that makes this story sting. The Wacom branding exists because Wacom did the right thing for decades. While other vendors shipped binary blobs, Wacom worked upstream — the xf86-input-wacom X driver, the kernel side, and crucially libwacom, the database that maps a device to its capabilities (button counts, pad layout, pressure ranges). That early, generous investment is why the org, the libraries and the data files all carry the name.

Fast-forward to today and those same repos are vendor-neutral in everything but their labels. The libwacom data directory carries entries for Dell, HP, Gaomon, Huion, XP-Pen and more. The companion wacom-hid-descriptors collection — the one KDE is actively asking users to feed device dumps into, two clicks, no coding required — is a general intake point for any tablet. The plumbing grew up and went multi-vendor; the signage never got repainted.

The rename has been debated for years and never happened, because renaming a load-bearing project is genuinely painful: import paths, packaging, distro metadata, documentation, muscle memory. Classic technical debt — cheap to create, expensive to service, easy to keep deferring until the day it blocks something concrete. That day arrived: a competitor's logo on a folder is now the stated reason four brands walked away from upstreaming. The debt finally came due, and the interest was a collaboration that should have been trivial.

How tablet support actually gets built now

For developers and the artists who depend on them, the genuinely good news is that the engineering path has gotten dramatically better — which is exactly why the branding stalemate is so frustrating.

Historically, non-Wacom support meant out-of-tree kernel modules. The DIGImend project carried that torch for years, reverse-engineering Huion, XP-Pen and similar generic tablets and feeding fixes toward the mainline hid-uclogic driver. It worked, but it was the slow road: every quirk meant a kernel patch, a release cycle, and a wait for it to trickle into distros.

The newer approach Revoy routes specs into — Red Hat's udev-hid-bpf, maintained by Peter Hutterer and Benjamin Tissoire — is a meaningfully different model. Instead of patching the kernel, you ship a small eBPF program that udev loads to fix up a device's broken HID report descriptor at plug-in time. No kernel rebuild, no waiting on the merge window. A quirk for a specific tablet can land as a self-contained BPF blob. For a category of hardware where every model has its own descriptor weirdness, that's the right tool: fast, targeted, and shippable independently of the kernel's glacial cadence.

The practical workflow if you want to help, or to get your own device working:

# capture what your tablet actually reports
sudo hid-recorder
# then submit the dump to wacom-hid-descriptors for a driver/quirk

The catch, and it's a big one: this still funnels through essentially two people. Revoy is candid that the Huion H610x, XP-Pen Deco 01 V3, Kamvas Pro 19 and Artist Pro 16/19 all work because Hutterer and Tissoire put in the effort, and that his pipeline stalls whenever they're unavailable. That's a brutal bus factor for an entire hardware category. eBPF fixed the mechanism; it didn't fix the fact that the whole thing rides on volunteer review bandwidth and a couple of Red Hat engineers' goodwill.

If you're buying a tablet for Linux

Strip away the politics and the buyer's calculus is unchanged, maybe even reinforced:

• Wacom remains the safe default. Decades of upstream work mean it generally Just Works in Krita, GIMP, Blender and under both X11 and Wayland. You pay more; you buy fewer surprises.
• Huion / XP-Pen / Gaomon / Ugee are increasingly viable — but support is reverse-engineered and lags new hardware. Buy the model someone already documented, not the one released last month. Check libwacom data and the descriptor repo before you spend money, not after.
• Avoid relying on vendor binary drivers. The proprietary Linux packages these brands ship are fragile across distro and Wayland changes; the FLOSS path, where it exists, is the one that survives upgrades.
• Your hid-recorder dump is a contribution. If you own an unsupported tablet, submitting its descriptor is the single most useful thing you can do for the next person who buys it.

The bottom line

This isn't a story about uncooperative Chinese vendors or precious trade secrets. It's a story about a name. The open-source tablet stack is technically ready for first-party collaboration — eBPF quirks, a shared device database, a clean intake process — and it's being held back by a logo on a folder and a maintainer pool of roughly two.

The fix is unglamorous and obvious: stand up a vendor-neutral umbrella, do the painful rename, and — far more important — fund full-time maintainers so the whole category doesn't hinge on spare evenings. Until then, Wacom keeps winning the Linux creative market by default, not because its competitors can't compete on Linux, but because nobody wants their name living in the rival's house. That's a terrible reason to leave good hardware unsupported.