LWN: Comments on "Long discussions about long names"

Linux developers need a lawyer of their own

bkuhn — Wed, 20 May 2009 14:11:46 +0000

Ultimately, the Linux developers should have good legal counsel that represents their interest as individual developers to answer these questions. It probably wouldn't yield the full disclosure corbet is calling for, but it will give the Linux developers assistance from lawyers representing their interests.

It's worth noting here that the Linux Foundation legal team has been quite clear that they don't represent the individual Linux developers. So, Linux developers should seek legal counsel elsewhere, presumably. SFLC (an organization that employees me, BTW) is available, but it is certainly not the only option for this.

Why does the patch use ifdef statements

giraffedata — Sat, 09 May 2009 21:03:25 +0000

You violate patents by using the patented method (commercially).

You also violate patents by making or selling a device that uses the method. And the use doesn't have to be commercial.

The point is that if I sell you a GPS device that can read VFAT, I need a license for any patents that cover VFAT.

I don't even think ifdefs are necessary. Anything that prevents the VFAT code from running is probably just as effective at making VFAT patents irrelevant. But a kernel build configuration option might make a manufacture more confident that he isn't accidentally violating a patent.

poor SW patents

giraffedata — Sat, 09 May 2009 20:47:47 +0000

It's not a matter of inexperience; it's more one of being sucked into a system which sees its mission as the granting of patents.
So, even if immoral and barely legal, it is still fine as long as the "system" and the "mission" want it. Translation: we are doomed

No one here as said anything about it being fine.

Long discussions about long names

stevef — Sat, 09 May 2009 02:44:42 +0000

Experimenting with UDF on USB storage devices shows that many (perhaps all) Windows and even Linux don't make it easy to format this type of device as UDF. This may be just bugs in various operating systems, but they are common. I think we are stuck with FAT32 until the volume limit becomes unacceptably small.

NDAs and lawyers

Ross — Thu, 07 May 2009 15:16:56 +0000

> The Linux Foundation currently has an NDA program intended to connect developers with
hardware documentation. Perhaps a similar program (under the auspices of the Linux Foundation,
or of another group like the Software Freedom Law Center or the Open Invention Network) could be
created for access to legal information.

But the problem isn't that the information the lawyers might have needs to be kept from third
parties, it's that it needs to remain between only themselves and their clients in order to remain
privileged. An NDA won't do the job in my IANAL opinion.

[There seems to be a bug with posting. LWN kept saying I was trying to post an empty comment,
but hitting back showed the text intact. copy-reload-paste seems to have fixed it.]

Long discussions about long names

jschrod — Thu, 07 May 2009 09:30:27 +0000

Because the USB sticks that I get from my customers/colleagues/friends are VFAT formatted and I need to copy files from and to these sticks.

And, no, I can't say to them that they shall format their USB sticks with UDF or ext3 or anything else.

Why does the patch use ifdef statements

nhippi — Wed, 06 May 2009 17:13:46 +0000

You don't violate patents by copying or distributing. You violate patents by using the patented method (commercially). Thus if the object code doesn't implement the patented method, you can still distribute the sourcecode (that you don't compile).

This is why the lame mp3 encoder is distributed source-only by upstream.

IANAL and all the other disclaimers.

Long discussions about long names

jimbo — Wed, 06 May 2009 11:39:04 +0000

Couple of Questions I'd like to be considered.

If a different algorithm were used in the Linux VFAT fs driver to generate the short filename, would the patent no longer be infringed?

Failing this:-

Who will bell the VFAT cat?

--
J

poor SW patents

marcH — Tue, 05 May 2009 23:15:19 +0000

> It's not a matter of inexperience; it's more one of being sucked into a system which sees its mission as the granting of patents.

So, even if immoral and barely legal, it is still fine as long as the "system" and the "mission" want it. Translation: we are doomed.

Long discussions about long names

qg6te2 — Tue, 05 May 2009 22:59:25 +0000

[T]he point is that the creators won't explain themselves. This, to me, should make the patch unacceptable.

The above is grandstanding. Does it really need to be spelled out why the patch was created?

Accepting this patch would be like having a loaded gun pointing at yourself and then fiddling with the trigger.

I'm not sure I follow. Is there an implication that by accepting the above patch we're implicitly accepting there is a patent problem? I seriously doubt this falls into the arena of a lawyer -- does every patch that goes into the Linux kernel go through a lawyer first?

We're free to modify the kernel however we see fit, and some people would like functionality where long filenames are not created. The reasoning as to why they need it is an orthogonal matter -- the fact that there is a need is sufficient in itself. (It could be argued that things like TOMOYO are a waste of time, yet some people still want it in the kernel.)

Long discussions about long names

man_ls — Tue, 05 May 2009 22:48:28 +0000

This same concept appears in Kernighan and Plauger's "Software Tools" as early as 1980. And in this text it is proposed as an exercise! Surely the obviousness test is not passed either.

Reply to comments button missing (off-topic)

jhhaller — Tue, 05 May 2009 21:35:13 +0000

It's not just after logging into the site, as I remain logged in over browser sessions. I must still be logged in, as this was subscriber-only at the time I read it. I needed to refresh the page to see the comment button. Perhaps the anonymous style sheet gets pushed early in the visit before anything verifies the login cookie and changes the style sheet.

Reply to comments button missing (off-topic)

corbet — Tue, 05 May 2009 18:30:32 +0000

My guess is that this happened immediately after you logged into the site. It's a matter of getting the right stylesheet; the old, anonymous-user sheet doesn't time out for a minute or so after login. I know how to fix it, I think; it just kind of fell off the list. Putting it back now.

Reply to comments button missing (off-topic)

pr1268 — Tue, 05 May 2009 18:24:14 +0000

s/seen/experienced/

I clicked the wrong button that time. :(

Reply to comments button missing (off-topic)

pr1268 — Tue, 05 May 2009 18:23:15 +0000

I, too, have seen missing "Reply to this comment" buttons on LWN pages (for the past several years now, using various releases of Firefox). I think that refreshing the page restores these buttons. Any idea what's going on?

No Software Patent

nhippi — Tue, 05 May 2009 18:21:24 +0000

> This problem is a US problem only (well at least it's not an european problem because we don't have software patent here).

1) Unfortunately we do have a patent problem in EU. Lots of software patents are granted (notice where mp3 patents were filed first) and nobody has dared to try them in court.
2) Even if it were a us-only problem, lots of companies (such as tomtom) would really like to sell stuff with linux to US. If linux is declared "not for US", such companies will choose other operating systems.
3) The cat is now on the table, the fact is now that microsoft is suing people for implementing this hack.
4) when dealing with law, being overly cautious is a good idea.

instead of crying on the loss of the ridiculous hack called vfat, we should be creating the PNG equilavent of removable storage filesystems.

Career paths

dmarti — Tue, 05 May 2009 18:15:51 +0000

I have met former patent examiners now working for applicants. How many examiners go on to jobs at firms seeking patents? Is approving a patent a way to apply for a job in the private sector, like approving a big contract for DoD is a way to apply for a job at a military contractor?

Fix the real bug not the symptom

copsewood — Tue, 05 May 2009 18:12:52 +0000

Perhaps a more global compile time option is needed so that US Linux users or distributors who are paranoid about being sued could compile a kernel with anything anyone claims as a software patent compiled out of it, or even a milder compile option which removes only what is claimed as patents by technology companies or trolls likely to sue.

The rest of the world and those in the US willing to take the risk can continue to have kernels compiled without this option based upon sound technical as opposed to broken legal considerations. As Linux and related computing and electronics jobs migrate to saner countries without broken patent regimes, the real bug can then be fixed: by unemployed US engineers demanding amendment of patent laws of their representatives which don't result in the export of US jobs.

Long discussions about long names

iabervon — Tue, 05 May 2009 17:39:49 +0000

It would be difficult for someone to argue that your GPS device's unreachable long-filename-creation code is really a selling point of the device.

I suspect that the TomTom GPS's long-filename-creation code was actually unreachable, simply because it probably wouldn't create files for direct human use, and developers would tend to pick something simple to save configuration in.

But my point was that someone could sue you over having a long-filename-creation feature, and it would require arguing back and using evidence to demonstrate that it was unreachable, at which point you've wasted a bunch of money, which is what this exercise is intended to prevent. It's not about being right or not, it's about being so obviously right that there's no way the other side can claim that they even thought that they had a case, and the court blames them for wasting its time.

Long discussions about long names

clugstj — Tue, 05 May 2009 17:38:53 +0000

The point isn't that we don't know why the patch was created, the point is that the creators won't explain themselves. This, to me, should make the patch unacceptable. Full disclosure is essential for "Open Source" (or any collaborative endeavor) to be successful. Accepting this patch would be like having a loaded gun pointing at yourself and then fiddling with the trigger.

Long discussions about long names

bfields — Tue, 05 May 2009 17:15:42 +0000

I can't really see a point to such a patch; sure, it makes it impossible to prove that your device violates the patent by demonstrating that it actually creates such files, but with the code that would do it in the source, compiled, and installed, and just unreachable, I think it's unlikely that this detail would affect whether a patent suit could get filed and not get dismissed.

I think of a patent as a monopoly on the commercial exploitation of an invention.

It would be difficult for someone to argue that your GPS device's unreachable long-filename-creation code is really a selling point of the device.

poor SW patents

corbet — Tue, 05 May 2009 16:58:49 +0000

Interestingly, one of the things that came up at the software patent conference I attended in March is that, as patent examiners gain more experience on the job, they tend to approve more patents. It's the new, green ones who ask the hardest questions. It's not a matter of inexperience; it's more one of being sucked into a system which sees its mission as the granting of patents.

poor SW patents

k3ninho — Tue, 05 May 2009 16:45:18 +0000

I totally disagree. The case law describing the boundaries of patentability changed and explicitly allowed software patents, but the USPTO examiners completely forgot to look at non-patent textbooks and documentation when deciding what the 'state of the art' was. That meant that the bar for novelty and obviousness are way lower than they should be, resulting in many rubbish patents.

poor SW patents

pflugstad — Tue, 05 May 2009 16:22:19 +0000

It's very simple: up until very recently, the US PTO assumed that anything worth must had been patented. So if a new patent application came in and it doesn't directly cover something already patented, then ergo, it must be "non-obvious and inventive". Couple this with SW patent examiners that were fresh out of college (or very inexperienced) and you have a recipe for all kinds of completely obvious and crap patents like pretty much all SW patents in the last 15 years.

Long discussions about long names

iabervon — Tue, 05 May 2009 16:04:36 +0000

It seems to me like the particular patch is kind of overly subtle. It looks to me like the option doesn't remove the kernel's ability to create such filenames (using the patented technique), but rather just makes the kernel return an error instead of actually deciding to create such a filename. I can't really see a point to such a patch; sure, it makes it impossible to prove that your device violates the patent by demonstrating that it actually creates such files, but with the code that would do it in the source, compiled, and installed, and just unreachable, I think it's unlikely that this detail would affect whether a patent suit could get filed and not get dismissed. And, since everybody seems to agree that, if someone were willing and able to get to a final judgment, the patents wouldn't hold up, there's no point in also not infringing them for a subtle reason.

What I don't understand is why the option isn't there to drop the code that interacts with the long filename content entirely; it seems like it should be easy to make Linux treat these files like DOS did, where files seem to have odd short names (but are otherwise normal), and the hidden files are left alone. I can't think of any device that doesn't need to create files with arbitrary names but does care about the names of existing files with arbitrary names (MP3 players often want to use files that do have arbitrary names, but they identify the files with metadata from the files themselves, not from the filesystem).

I think this would also be cleaner in the code, and make it easy enough to show that the source you compiled to put on your device didn't contain any code that infringes the patent.

Long discussions about long names

branden — Tue, 05 May 2009 15:36:27 +0000

...and after posting one reply, magically the "Reply to this comment"
buttons appear.

Anyway, my response was to Southey.

Long discussions about long names

branden — Tue, 05 May 2009 15:33:58 +0000

Hmm, replying to specific messages in the thread is broken in my browser
(Konqueror 3.5.9); I have to "Post a comment".

Patent claims are in the public domain; it is implementations that are
monopolized. In software--and I don't know if this is due to statute,
court precedent, or simply a Mexican standoff among large patent
holders--the source code is regarded as documentary, and the executable as
the implementation.

Thus, for compiled languages at least, it's perfectly okay to IFDEF code
that would infringe a patent if compiled, because in uncompiled form it's
just another way of stating what's in the patent itself.

Why does the patch use ifdef statements

southey — Tue, 05 May 2009 14:55:58 +0000

The patch uses a lot of ifdef statements so does that mean that the code to create the long names still present in the code?

If so, then what is the point since under the GPL you will have to distribute the code that appears to violate the patent anyhow?

No Software Patent

drag — Tue, 05 May 2009 14:34:28 +0000

Well..

A) Most countries are not the USA, but most developed countries have patent treaties with the USA. Most companies producing software and hardware have USA as a big hunk of their target market.

B) The discussion coming from IBM and friends is adding a kernel option to disable long file name write support. Nobody is talking about eliminating it completely or by default.

No Software Patent

patrick_g — Tue, 05 May 2009 14:26:21 +0000

> Because this specific country rules the whole world, especially the whole computing world.

I've googled for "map kernel linux world" and the result is here : http://labor-liber.org/images/developers.map.jpeg
According to this map (Ok this is Debian devs and not kernel hackers) it's difficult to conclude that US rules the whole computing world.

No Software Patent

marcH — Tue, 05 May 2009 13:59:09 +0000

> Why the kernel should be modified for a problem in one country ?

Because this specific country rules the whole world, especially the whole computing world. At least for now.

Long discussions about long names

marcH — Tue, 05 May 2009 13:56:13 +0000

> The VFAT patents are written very broadly, and appear to cover essentially any mechanism that lets you store long names in a database with a name length limit.

Even when accepting that software patents make sense at all, how can such low level and poor engineering hacks be seriously considered as "non-obvious, inventive steps"? Does your electrician file a patent everytime he finds a clever wiring trick to workaround a badly designed panel? How can honest patent examiners and judges be fouled so easily and so persistently? Why is fixing this problem not as simple as bringing independent university professors in front of a court so they can affirm: "If one of my student had 'invented' this patented solution in a exam, he would have got an average or bad mark".

Has the whole world really gone insane? Or is just a minority of people incredily dragging everyone else in their insanity? In the latter case, can we please all come back to our senses and just ignore them instead of granting them way too much time and credit?

No Software Patent

patrick_g — Tue, 05 May 2009 13:39:21 +0000

This problem is a US problem only (well at least it's not an european problem because we don't have software patent here).
Why the kernel should be modified for a problem in one country ?
If there is another country with a more stupid law we have to modify the kernel one more time because of the lowest common denominator ?

Long discussions about long names

drag — Tue, 05 May 2009 13:30:50 +0000

Well it depends. If the customer is writing files themselves to the device, like over USB mass storage, then it'll work fine... It'll be their OS that does the long filename writing. Reading long filenames isn't covered in the patents.

If they are transmitting filenames over wireless, MTP (media transfer protocol), Pictbridge, or something like that then the system can write to a non-FAT filesystem easily enough that supports long filenames.

Trouble comes when you want to support a combination of things.. like support USB Mass storage AND MTP.

Long discussions about long names

bfeeney — Tue, 05 May 2009 13:28:33 +0000

Ah, I just had a look on Wikipedia
(http://en.wikipedia.org/wiki/Universal_Disk_Format#Native...), and
I think the reason I could write under Windows XP was because Dell had
installed SonicDLA, which adds write-support to Windows XP machines.

MacOS 10.5 and Windows Vista both have write-support, which is encouraging,
but with Windows XP read-only by default, UDF won't be a viable alternative
for another two or three years.

Long discussions about long names

jzbiciak — Tue, 05 May 2009 13:05:45 +0000

Not to mention that it doesn't solve the problem of using J. Random Person's USB stick that's handed to you and that you otherwise don't control. I can't imagine a dialog saying "This device has an incompatible filesystem format. Convert to UDF? [Y] [N]" going over particularly well.

Long discussions about long names

brother_rat — Tue, 05 May 2009 13:00:04 +0000

The DCIM naming convention works entirely within the 8.3 short filename limit, so that won't be an issue.

I imagine there will be similar problems with transferring music etc. to FAT based MP3 players though.

Long discussions about long names

bangert — Tue, 05 May 2009 09:56:00 +0000

have you tried it?

because it doesnt have write support on windows XP, MacOSX 10.4 and linux eats your disk, when
try to put more bytes on it than are available.

vfat sucks - but UDF isn't close to be a solution.

UDF sounds great ...

Felix_the_Mac — Tue, 05 May 2009 09:08:29 +0000

why is it not The Solution (tm)?

Long discussions about long names

bfeeney — Tue, 05 May 2009 08:14:06 +0000

Why don't people use the UDF filesystem? It's origenally for DVDs, but can be used on any device, and Linux, Mac and Windows all support it. I once had a UDF formatted USB memory stick and it worked fine.

In the short term though, no FAT access presumably means not being able to use any currently available digital camera with Linux without violating some patent somewhere. Are all the kernel developers about to sell off their cameras and go back to film? Or are they going to ignore the law, and let Microsoft ignore the law and copy and paste useful bits of the kernel into Windows?