| |||||||||||||
|
| |||||||||||||
On corporate PR and proper credit
On August 2, MontaVista issued a
press release claiming that its new hard realtime kernel "breaks
barriers" and enables Linux to be used in hard realtime applications. The
PR includes a link to the realtime Linux page on the
MontaVista site which furthers these claims, showing how its latest kernel
has much better interrupt and scheduling latency than the mainline 2.6
kernel. All of this, we are led to believe, is MontaVista's work. There
are, however, no pointers to the actual code in question. When we asked
MontaVista where the code could be found, we got a bit of a surprise.
MontaVista's PR company sent us a pointer to Ingo Molnar's realtime preemption patches. These patches have been covered several times in the past on LWN's kernel page; look under "realtime" in the LWN Kernel Index for the articles. Ingo is an employee of Red Hat, not MontaVista, and the realtime preemption patches are mostly his work. So it would only be proper for MontaVista to credit that work in its PR materials. Instead, we see claims like:
MontaVista's new hard real-time developments filter the roughly 6
million lines of Linux code down to around 100 critical
interrupt-code segments.
(from the press release) or:
MontaVista's most recent work further enhances the
community-established real-time foundation, which pushes the Linux
kernel's performance and predictability. Today's real-time
milestone further advances the kernel toward predictable response
times, making this the last hurdle in evolving a hard real-time
Linux kernel.
(from the realtime Linux product page). Ingo Molnar's name is not mentioned anywhere on MontaVista's web site or in the press release. MontaVista is not entirely without credit in this work. A patch posted by MontaVista's Sven-Thorsten Dietrich in October, 2004 integrated much work done by others, and included the important idea of replacing kernel spinlocks with a semaphore-like mutex type. Two days after that posting, however, Ingo Molnar came out with his first realtime preemption patch which, while using many of the ideas from Sven-Thorsten's patch, was a new implementation with many new ideas. The patch has evolved considerably since then, and bears little resemblance to its early version, much less the MontaVista patch. It includes work contributed by others at Red Hat, IBM, LynuxWorks, and elsewhere. There is nothing wrong with MontaVista taking the realtime preemption patches and turning them into a product. That is one of the freedoms afforded by free software. MontaVista can rightly claim credit for some of the ideas found in that patch set, and, perhaps more importantly, for giving a kick which helped to get the whole process going. But when MontaVista claims exclusive credit, and does not even see fit to name the person who has done the lion's share of the work, it is pushing the system a little too hard. Credit is an important currency in the free software community, and it is such an easy currency to hand out. MontaVista's customers know that the company did not develop the entire Linux system; they will not care that the realtime preemption code, like most of the rest of the system, came from somewhere else. So there is no reason for MontaVista to claim credit for the work of others in this way. If MontaVista wants to participate in the Linux development community (and its participation would be welcome), it would be well advised to send out a followup release giving proper credit for the code it is shipping. Update: the realtime products page now has (at the end) a list of contributors to the realtime preemption patch. MontaVista should also be credited for the interrupt patch posted by Daniel Walker in June and subsequently added to the realtime preemption patch set. (Log in to post comments)
On corporate PR and proper credit Posted Aug 2, 2005 21:29 UTC (Tue) by hamjudo (subscriber, #363) [Link] At first corporate PR was just directed at consumers. These days, in large corporations, the PR department also has to keep in mind how the PR will affect the company's own employees. Likewise, for publicly traded corporations, the reaction of the shareholders matter a great deal.Companies that deal with Open Source, have to pay attention to the developer community. Most PR departments will learn, but some won't and will destroy their own companies in the process. The smart PR departments will incorporate community feedback into their marketing processes, which will lead to higher quality, less embarrassing press releases.
Just a hobby, won't be big and professional ... Posted Aug 2, 2005 21:33 UTC (Tue) by brugolsky (✭ supporter ✭, #28) [Link] Montavista ought to lay claim to integration, testing, QA, and support of Ingo's efforts. And they could emphasize their initial contributions to preemption in the 2.6 kernel. But to omit mention that the work is 80+% Ingo's is simply unethical.
Unfortunately, we are seeing more companies that enhance and/or bundle free software issuing these type of press releases. There's a strong desire to portray themselves as "big and professional," in contrast to "a handful of hackers with a hobby." They don't get it -- Linux has bested the "big and professional" operating systems -- in no small part due to the major contributions of a handful of individuals like Ingo Molnar, whether working in their spare time, or now employed to do Linux development full-time.
Ingo Molnar and the many virtues of parsimonious design. Posted Aug 2, 2005 22:24 UTC (Tue) by brugolsky (✭ supporter ✭, #28) [Link] Fair reader, if you'd like to wash that bad taste from your mouth, and re-live the thrill brought on by Ingo's shredding of the MindCraft benchmarks, I'd suggest (re-)reading Timothy Dyck's fine 2001 article on the TUX web server, TUX: Built for Speed.This amusing sentence appears in a companion article that details the eWeek benchmarks ( Devils and Details of Benchmark Tests): "The most unexpected testing issue that came up was our inability to saturate the Tux-based Web server on our standard test server configuration because it was so fast, a problem that forced us to remove two processors from the four-way Dell PowerEdge 6400 server (which was equipped with two Gigabit NICs and 2GB of RAM). This was the only way the testbed of 80 workstations could max out the Tux server (see benchmark chart)."
Advertising clause Posted Aug 2, 2005 21:41 UTC (Tue) by rfunk (subscriber, #4054) [Link] I'm sure this is exactly the sort of thing that the original BSDlicense's inconvenient "advertising clause" was intended to prevent. (Of course, it caused problems of its own.)
On corporate PR and proper credit Posted Aug 2, 2005 21:58 UTC (Tue) by drag (subscriber, #31333) [Link]
There was some conflicts a while ago with Linux kernel developement and "real" realtime capabilities being brought into Linux.
Montavista and another company (maybe windriver?) had developed extensive modifications with the Linux kernel to make it realtime-capable.
However they were not acceptable to Linus based on that they were very intrusive to the kernel and had code spread all thru out the system. This caused some friction...
Ingor was working on 'volentary preemption' stuff for a long time, I think, but when montevista and friends were fighting to get their mods into the kernel he went and started developing realtime-preempt heavily in a hope to get the modifications integrated into the kernel in the distant future buy doing the mods in a Linux-coding standards acceptable way.
I am not sure about this, but this is the general impression that I've got. MonteVista maybe using a lot of Ignor's code, but they have put a lot of work into the kernel themselves even if that work wasn't very accaptable to Linus ultimately.
I don't know if this is a very serious issue.
On corporate PR and proper credit Posted Aug 2, 2005 23:55 UTC (Tue) by roelofs (subscriber, #2599) [Link] Montavista and another company (maybe windriver?) had developed extensive modifications with the Linux kernel to make it realtime-capable.I believe the other company was Timesys (mentioned near the end of the article), not WindRiver. The latter is a recent convert and has not, to my knowledge, contributed much to Linux. Greg
On corporate PR and proper credit Posted Aug 5, 2005 18:15 UTC (Fri) by piggy (subscriber, #18693) [Link] Part of my job is making sure that TimeSys continues to be active in the Linux kernel community.
FWIW, I've got about 80 submitted TimeSys kernel patches in my tracker. Not all have been accepted and most are admittedly minor, but we have been pretty active. We don't track our kernel patch submissions any more, so I can't give current data.
Scott Wood of TimeSys wrote the soft IRQ threads implementation in Ingo's patch set.
Jon Cooper is a TimeSys employee who is working full time with Ingo.
TimeSys is active, but we've been keeping a low profile. We're much more interested in seeing the work completed and accepted than in claiming that we did it.
On corporate PR and proper credit Posted Aug 2, 2005 22:21 UTC (Tue) by sbergman27 (subscriber, #10767) [Link] From the bottom of their realtime products page:
------
Some key contributors to the overall real-time effort include (alphabetically):
* Ingo Molnar (maintainer)
* Esben Nielsen
On corporate PR and proper credit Posted Aug 2, 2005 22:24 UTC (Tue) by corbet (editor, #1) [Link] Interesting. I don't believe that was there before - I looked hard for it. Maybe some of the grief I gave their PR person carried through. We need an internet archive with sub-hour resolution...:)This is a step in the right direction anyway, and a good thing.
On corporate PR and proper credit Posted Aug 2, 2005 22:47 UTC (Tue) by sbergman27 (subscriber, #10767) [Link] I imagine it wasn't there. I'm going to guess that 90% of PR people hired by OSS companies require "orientation" regarding how things work in OSS, and what is considered acceptable, and what is considered an inexcusable insult.
Honest mistake, I suspect.
Less than two hours to fix the problem after it's pointed out is not too bad.
If Montavista's security advisory response time is this good... Posted Aug 3, 2005 0:45 UTC (Wed) by xoddam (subscriber, #2322) [Link] ...then I'd consider them a first-preference vendor.
Thanks Jon for raising the bug, and thanks to Montavista's publicity people for correcting the error so quickly.
(I guess the text of a web page doesn't need quite as much integration and testing before deployment as, say, a major rewrite of the kernel's synchronisation primitives :-)
On corporate PR and proper credit Posted Aug 3, 2005 8:29 UTC (Wed) by lamikr (guest, #2289) [Link] I think this situation has showed how open source software works in it is best not only in the producing code but also for documenting changes and following history.
0) Some people inspired Montavista's Sven-Thorsten Dietrich to submit rt patch.
On corporate PR and proper credit Posted Aug 3, 2005 11:54 UTC (Wed) by mingo (subscriber, #31122) [Link] i'd like to reflect to points #0 and #1:
> 0) Some people inspired Montavista's Sven-Thorsten Dietrich to submit rt
there were 2 days between the two announcements, and my PREEMPT_RT work has grown out of the "Voluntary Preemption" patches in a process that lasted weeks and months, not two days. So the original patch includes no code and no ideas from MontaVista's original patch simply due to there not being much time for anything than a quick look at MontaVista's code to be able to comment on it.
but despite this somewhat confusing and overlapping initial start, healthy cooperation started between the two projects pretty early on, via the universal communication method of sending me kernel patches ;-) MontaVista sent a couple of useful pieces of their patch to me and is generally one of the top contributors to the PREEMPT_RT patch.
|
|
||||||||||||