Joerg Schilling
In a nutshell
Joerg maintains libscg, a cross-platform SCSI library, and cdrtools, a suite of CD/DVD writing tools.
Joerg willingly ignores, misinterprets, or makes shit up to avoid any information he doesn't like.
Joerg doesn't like any information which introduces changes to any of his personal conceptualizations, including those of software, people, entities, licenses, and himself.
Pictures
Joke fodder
thinks every device that is capable of accepting any subset of SCSI commands should be usable over a single kernel-emulated SCSI transport. http://lkml.org/lkml/2006/1/31/47
says every emulated SCSI transport should be SCSI-enumerated (for example: 2,0,6), even going so far as to specifically claim that libscg should be able to enumerate an iSCSI "bus" by means of a kernel-provided mechanism to scan the "bus" (which is, of course, an IP network). http://lkml.org/lkml/2006/1/31/83
- has sweepingly called Linux and FreeBSD kernel developers incompetent and accused many of them of conspiring against him to break libscg. (from "Joerg Schilling flames Linux on blog" thread, or possibly the blog post itself. I can't remember now.)
- says that known kernel security problems should have fixes delayed for six months so that non-essential (trivial, even?) user-space program authors can be consulted. (from "Joerg Schilling flames Linux on blog" thread)
- says his SCSI implementation in Solaris is better than the SCSI implementation in Linux because his hasn't changed significantly since August of 1986. (from "Joerg Schilling flames Linux on blog" thread)
Doesn't really understand said implementation (or maybe the argument?) http://lkml.org/lkml/2006/2/13/249 (to clarify, the host,lun,id fields total 32 bits. the id field is much, much smaller.)
seems to have no conception that Unix kernels can undergo constant development. http://www.ussg.iu.edu/hypermail/linux/kernel/0408.3/0129.html
- made his own "definitive" SCSI implementation patch for Linux (2.2, I think?) which he completely failed to maintain. He then blamed Linux developers for it being broken within a week.
- has a poor rationalization of hotplug in general, and absolutely no conception of how Linux hotplug works.
- has no idea what udev and sysfs are, and accuses them of working with the kernel to hide information from him. (from "Joerg Schilling flames Linux on blog")
uses SCSI enumeration specifically because he thinks it's stable across resets and hotplug events (it's not. Not even under Solaris) and that device nodes can't possibly be stable to such a degree (which they are with udev). http://lkml.org/lkml/2006/2/10/83 http://lkml.org/lkml/2006/2/9/186 http://lkml.org/lkml/2006/2/13/189
keeps -scanbus around because he thinks users want it. Sort of how he thinks users prefer 102,395,190 to /dev/cdrw. http://lkml.org/lkml/2006/1/26/53
expects DMA support for SCSI-emulated buses, but SCSI does not support DMA. http://lkml.org/lkml/2006/2/13/363
- Experienced Linux kernel developers hate and shun discussions with Joerg. They've learned.
- Joerg is from Berlin and is not very good with English. Especially when excited.
does not accept patches to cdrecord (a GPL program) based on the fact that they are licensed under the GPL. http://lkml.org/lkml/2006/2/3/204 http://lkml.org/lkml/2006/2/6/161
actually has no idea how the GPL works; hates Redhat for applying Linux-specific patches http://lists.debian.org/cdwrite/2004/03/msg00091.html http://lists.debian.org/cdwrite/2003/11/msg00181.html http://lists.debian.org/cdwrite/2003/11/msg00190.html http://lists.debian.org/cdwrite/2003/11/msg00195.html
hates Debian for applying Linux-specific patches, further hates them for taking over user feedback provisions on tainted versions (something he'd only bitch about anyway: see previous entry) http://lists.debian.org/cdwrite/2006/01/msg00145.html
- hates Suse for applying Linux-specific patches.
selectively ignores parts of conversations that he doesn't like, going so far as to loudly claim that those messages never happened at all. http://lkml.org/lkml/2006/2/13/209 and a few less clear-cut others
is paranoid that Debian closes fixed bugs on Debian-patched versions of his software, claiming that they are hiding evidence. http://lkml.org/lkml/2006/2/9/149 http://lkml.org/lkml/2006/2/14/302
- Thinks the GPL is a bunch of crap because nobody ever provides useful patches to him and people constantly make "unauthorized" branches. Reminiscient of the braindead insularity of the Nessus folks.
Thinks the GPL does not apply to binaries (Perhaps thinking that binaries are not a derived work)? http://lists.debian.org/debian-devel/2006/08/msg00569.html
Thinks the GPL and CDDL are compatible, despite the creators of both the GPL (Eben Moglen) and the CDDL (Simon Phipps and Danese Cooper) saying that this is not the case. When shown video of CDDL drafters making this explicit, Joerg claims that they were either mistaken or coerced based on the facial expression of someone who is actually offscreen. This was a Debian event that Joerg did not actually attend such that he could know anything about said expression. http://lists.debian.org/debian-devel/2006/08/msg00563.html
Repeatedly fails to bother testing his own lkml bug report. Similar to ESR's asinine iPod claim. http://lkml.org/lkml/2006/10/17/132
http://www.fooishbar.org/blog/tech/joerg-2006-02-21-12-47.html - autoconf outdated (This is almost certainly a legitimate argument on Joerg's part. Autoconf has some serious problems, and Joerg is possibly the most likely person in the entire universe to need significant workarounds. Though, admittedly, I don't know anything about the quality of his replacement. The "smake" and "star" rigamarole could, however, lead to jokes of him adding "s" to basic GNU tools ad absurdum and claiming they're more POSIX compliant and portable. -JoeRayhawk) (My point was that he was saying the only way to be portable, was to rely on a make implementation that no-one ships, which is laughably stupid, even compared to all the background noise. autoconf is certainly abysmal, sure, but that wasn't the point. -DanielStone, author of said entry)
(Joerg speaks of technical considerations rather than practical ones. It seems rather Schillingish to judge his critique as if it had been written in terms of a value system he had had no intention of using (Or, in his case, ever comprehending). I suspect Joerg is psychotically incapable of seeing that misunderstanding, whereas I suspect you are sane enough, but simply choosing not to. If it's for comedic effect, you're working on a level only five other people will comprehend, and I encourage you to write for ELER. -JoeRayhawk)
More specific amusements:
http://lkml.org/lkml/2006/2/14/252 "Surely you're not suggesting that TOG's job is to rubber-stamp bad Solaris decisions..." -- H. Peter Anvin
http://lkml.org/lkml/2006/2/13/192 "Talk about people who "do[...] not know how things work on Linux" and name /etc/vfstab in the same breath. Brilliant!" -- Matthias Andree
http://lkml.org/lkml/2006/10/19/102 "I can't think of a person I'd most rather have as lobbyist for a rival OS." -- Alan Cox [sic]
Analysis
I believe the motivation for all this is that Joerg wants libscg to be perfect and beautiful cross-platform code. However, rather than implementing a superset of all features and only using what the underlying system hasn't already implemented (like, say, libasound does with hardware), Joerg expects all systems to conform to his (feature subset) conception of what beautiful libscg should be. To this end, he viciously hates outside patches and operating-system specific code to the point of accusing every major distribution of breaking cdrecord/libscg by patching them. When he is forced into operating-system specific support, he throws very childish tantrums for the world to see. Cdrecord using dev nodes with warnings under Linux, but not under Solaris, is a good example.
To a certain extent, some of this makes sense in a roundabout sort of way. Joerg's conceptions of Unix date back twenty years. The IDE bus would be an afterthought for mainframes and high-performance hardware, so half-assedly allowing access to ATAPI or even all ATA with very well understood SCSI code and interfaces seems like a reasonable plan. Meanwhile, with modern free kernels (used largely on low-end consumer machines and embedded systems) IDE is a staple interface. Requiring SCSI transport support in a low-memory environment for the majority of systems which don't even have it seems like a really stupid idea. His feature stability and security expectations also make (*cough* a limited amount of *cough*) sense under ye olde Unixe expectations.
His discussions with new-Unix kernel developers, such as FreeBSD and Linux, are a constant fight against progress which are hilarious to read because he refuses to comprehend or sometimes even acknowledge new or changing interfaces and features. When he's simultaneously calling developers incompetent and trying to tell them how to do their job, most of his supremely helpful suggestions involve moving kernel development backward several years.
Here's an RSS feed for people into that sort of thing. He has two-week-long flame-wars every six months. http://lkml.org/groupie.php?author=Joerg+Schilling