Making Sure Ubuntu Works with Hardware

The Ubuntu Friendly program, much like the old Fedora data gathering and other similar programs, aims to find out what hardware Ubuntu works on and which hardware causes issues. This is a very important part of out ability to assess both regressions and entrant hardware comparability and I commend anyone’s participation in the program as vital to Ubuntu’s future progress.

I do however want to add something to the discussion about hardware testing. I think one of the things we get wrong when gathering data about hardware is that we have no clue how much hardware is actually out there, we indeed can not be sure what it is possible to test with and what we’ve failed to test. And this half-data doesn’t just impact hardware testing…

A classic example from our community is the guttenprint driver and open printing website. The website lists all the printers which other users have added to the database. This means that there are inconsistencies in naming, duplicate entries and lots and lots of missing entries. This doesn’t help developers, they can’t know for sure how good the coverage of the guttenprint driver is looking at the website.

The solution for gutten print and I think for projects like the Ubuntu friendly program is to spend some time researching and documenting the hardware-space; i.e. writing down all possible published hardware. A good solid coverage list can turn the seemingly infinite task of say getting a picture for each hardware into a finite game where you win when all known hardware has a picture.

This is the same as testing, as sure as it is the same with producing printer drivers. We should know for sure what our possible scope is before passing over the torch to user data. Users are not going to be as bothered about testing their computer if the message is: “Come test your computer, it might be possible your data is unique”, but they could be much more motivated if the message is targeted as: “This type of computer has never been tested and you’d be helping Ubuntu out a great deal if you can run the test suite.”

Not just good for targeting, but also good for testing progress. We can see how well our testing programs are going, how much coverage we can produce in time for the release or RC. It would be great to have a competition between LoCos to see which group can dig up new and unique hardware to test.

So with all this in mind, I appeal to all those involved in the Ubuntu Friendly program to consider doing the diligent thing of going out there to OEM websites and documenting everything you can find. It doesn’t have to be in depth, we might not even need to know what the hardware is for sure, but it would be great to be able to use smart tactics.

What are your thoughts dear reader?

Tags: , , ,

7 Responses to “Making Sure Ubuntu Works with Hardware”

  1. Sandman says:

    How are this going to be organized? I am not against this idea.. it sounds awesome! But without a good back-end a lot of work will go lost. It need to be well categorized but easy enough for ppl to bother. A well defined check-list of necessary components that can be used. A rating system on how well documented one hardware is.

  2. ethana2 says:

    I would also like to be able to make a list in such a site of all the hardware I own, so if a bug affects a driver, you can *instantly* see how many people it does or may affect without involving everyone in the bug report. To a degree I think this could be done automatically.

  3. studentz says:

    Playing with nxt Lego Mindstorm I found that the natty bluetooth manager does not let me connect to the device. After installing blueman, which showed me the serial option, everything was smooth. Here the problem is worse, it is not the kernel, it is not hardware compatibility, it is a deficient front end app. What a lame thing :(

  4. Jussi says:

    Have you thought about approaching manufacturers for a list of products? Most manufacturers keep such a list, and if approached in the right way, would probably share such a list.

    We need to see if we can work with the manufacturers more, advertise to them the benefits of sharing such information – they are getting free work done after all, and can then claim compatibility with ubuntu etc.

    Of course, not all manufacturers are going to share, and not all manufacturers exist anymore, so it wont be perfect, but a sight better than it was.

  5. Marc Tardif says:

    The Ubuntu Friendly program uses the Results Tracker to get test results about hardware in the community. The site makes a best effort to answer questions like “how much hardware it actually out there.” This is made possible by separating the concepts of models (any Latitude 2120), units (my 2120) and even states (my 2120 on a particular date). However, this level of granularity does not transpire to the Ubuntu Friendly website to keep the site… friendly.

    Furthermore, you mention the concern of “inconsistencies in naming”. In order to avoid this problem, the Results Tracker also makes a best effort to avoid exposing naming to human error. This is accomplished by gathering as much data as possible directly from the hardware which is not without its own problems. For example, some manufacturers forget to set some BIOS data and you might be pleased to know that Dohickey was used as an inspiration to blacklist some of these occurrences. Thanks!

  6. doctormo says:

    Marc – Ah yes the classic ‘can’t be bothered to fill in the bios’ problem. Only the user can help there unfortunately.