Date: Fri, 26 Mar 1999 02:34:14 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: wes@softweyr.com (Wes Peters) Cc: tlambert@primenet.com, brett@peloton.physics.montana.edu, brett@lariat.org, jkh@zippy.cdrom.com, advocacy@freebsd.org Subject: Re: FreeBSD emulation for linux Message-ID: <199903260234.TAA01443@usr09.primenet.com> In-Reply-To: <36F9D424.2397F563@softweyr.com> from "Wes Peters" at Mar 24, 99 11:13:56 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > > By getting the vendors to count it. What this has to do with a > > > FreeBSD emulator in Linux is completely beyond me. > > > > You have to subtract out FreeBSD users using Linux binaries from > > the FreeBSD native binary market size. > > No, you use the count of the FreeBSD users using Linux binaries to > convince them it's time to support a native FreeBSD version if the > numbers warrant it. I thought that was obvious. It *is* obvious > to everyone on the planet but you and Brett Glass. As someone who has been in that position, you're not convincing me. Century Software developed Xenix releases on Altos hardware specifically because the resulting binaries would run on SCO Xenix, Altos Xenix, Intel Xenix, and Cubix System V. The SKU was "IA" for "Intel, Altos". If I'm a software vendor, I'm going to minimize the number of SKU's I have to deal with. Period. If nothing else, I can leverage the unit production back and forth over a number of target platforms without having to nail the unit numbers I expect on a specific platform. At best, I cut my porting overhead and reduce my development (and validation) costs. The best that you are ever going to get out of a software vendor is them eating the validation overhead on your platform and (assuming validation actually passes with no work on the part of the vendor aside from running the tests), "also-runs-on" branding on a sticker on the outside of the shrink-wrap. If I were in that position, that's the most you're going to get out of me -- and I'm a FreeBSD advocate. > > > This is exactly what I disagree with; the OS/2 analogy is not apt > > > regardless of the source. OS/2 died for a number of reasons, not > > > the least of which was that IBM marketed it horribly and delivered > > > a non-working system for the first several major releases. > > > > Let's divorce that little "and" as "necessary but not sufficient". > > So your posit is that FreeBSD is being marketed better than OS/2 was? > > My posit is that FreeBSD hasn't been marketed at all yet, so we cannot > yet tell if it will be successful. The differences between FreeBSD > and OS/2 are so large I don't think there are any lessons to be learned > there at all. There are always lessons to be learned. One lesson that should be foremost when we're talking about FreeBSD running Linux binaries is that unless they run without having to be pounded on, then defacto, FreeBSD is delivering "a non-working system". Just as OS/2 was delivered. I disagree with you about whether FreeBSD has been marketed or not, as well. Jordan is implicitly marketing FreeBSD as "something to run your Linux binaries". He's reinforcing this message from a number of angles. This may not be exactly what he's saying, but it's what at least some of us are hearing -- and again, we're FreeBSD advocates. How much worse must the message be for non-advocates, or for people who are hearing about FreeBSD for the first time during Jordan's speech about "if you are a software vendor, port to Linux". The message isn't focussed solely on software vendors; it's given at places like the O'Reilly Open Source Forum... it's given in front of press people. > FreeBSD marketing is starting to pick up the pace, but I worry about > that personally. I've never needed much in the way of commercial > applications, and am pretty much happy to wait for developments like > GNOME and AbiWord, and support whoever I need to make sure they work > well on FreeBSD as they reach maturity. I definitely worry about FreeBSD marketing. But so long as there is going to be a FreeBSD presence, then it might as well have the most positive possible spin on it. If Jordan is going to show up and tell software vendors to port to Linux (yes, I realize he qualifies this as "software vendors entering the UNIX market"; software vendors, however, don't hear that part), then someone needs to spin it properly. One way to do this would be to replace "Linux" with "FreeBSD" in the statements Jordan makes; but to do that, FreeBSD binaries would have to make the same promise that Jordan is making on behalf of Linux binaries: "port on X, run on X and Y". > > > I personally think a FreeBSD emulation layer for Linux is a huge waste > > > of time, because I strongly doubt you will get Linus to put it in the > > > kernel, nor will you ever convince any of the major distributors -- > > > i.e. RedHat and nobody else these days -- to put it in their distribution. [ ... ] > And I still doubt you'll get them to ship a FreeBSD compatibility module. > I see them being interested in that about 43.7 decades after hell freezes > over. Well, you now have a wonderful opportunity for an "I told you so", predicated upon someone providing a FreeBSD compatability module for Linux. Regardless, it's irrelevent what *ships*, it only matters what's *available* so that the story you tell is true. > > I don't understand your statement. Why would the Solaris and SCO > > employees dwho made the decision to write the Linux ABI code be > > running FreeBSD? > > Because FreeBSD is a better system than SCO, Solaris x86, or Linux. For > just about anything you care to name. If they were picking the ABI on > technical merit, they'd obviously pick the FreeBSD API to run on, due > to the quality of the system and the entirely reasonable price. Since > the decision is not being made for reasons of technical merit, any > discussion of technical merit is a moot point. I disagree. Solaris (and UnixWare) have a number of technical points on which they still far outstrip FreeBSD. You're forgiven for your advocacy, though; everyone has their religion. > > > So, what do you do? Get some talented FreeBSD hackers willing to keep > > > up with the shifting sands of the Linux ABI as well as possible, point > > > out to people that most Linux apps run on FreeBSD too, so they're not > > > risking much, and do an effective job of advocacy in getting the vendors > > > to decide which side of the open source bread the butter is on. > > > > This limits you to a zero sum game the size of the Linux market. > > Geez you're being obtuse. It limits you to the size of the Linux + > FreeBSD + anything else that will run the Linux ABI market. Since that > seems to be the ABI of choice these days, you're attempting to play into > the largest UNIX application market possible, not some half-baked crock > that doesn't exist yet. It limits *you* (FreeBSD), not "you the software vendor". > > But it matters a hell of a lot to FreeBSD whether the app > > is a Linux or FreeBSD binary, in terms of gaining critical > > mass > > No, it doesn't, not if they're willing to say it's a Linux *and* > FreeBSD app. That's been the point of this entire argument, which > you've consistently missed, so now I'll ram it down your throat: > It doesn't matter one damn bit what the bits in the ELF header say, > because most users cannot tell the difference between one and the > other, it matters only that it runs on FreeBSD and says it runs on > FreeBSD on the fucking box. > > Do you get it now? I understand. Now you are chasing the Linux ABI problem, which was my point: the Linux ABI is fluid enough that you will have a hard time saying that about two different Linux ditributions, let alone about a FreeBSD implementing the ABI from a particular Linux distribution. The problem is as hard as tracking Win32 unde OS/2. Look at how long Linux threaded applications failed to run on FreeBSD. > > And it matters a hell of a lot to FreeBSD whether the app > > is running on a Linux or a FreeBSD system. > > And so what we're trying to do is find the highest-percentage shot at > getting those applications to run reliably on FreeBSD, and to mention > FreeBSD in the materials advertising the applications. OK. You pursue that goal. Brett and I will pursue getting FreeBSD native binaries, as a seperate goal. > Which do you think is more likely to happen: asking vendors to port > their working Linux apps to some highly speculative FreeBSD emulator > for Linux, or basically handing them a done deal for another 1.5 > million potential users with ZERO effort required on their part, other > than sticking a logo we supply them with on their web page, product > box, etc? Which would YOU do? I've told you what I, as a vendor, would do (*did*, in fact), above. > So far, we've got ONE. They're waiting for US to deliver the required > marketing materials, and I'm working on that, with assistance from > several others here. Of course, so far we've asked exactly ONE, so > we're hitting 100% so far. > > Why don't YOU go ask someone? It can't hurt, the worst they can do is > say "No." They can't take away your birthday. I mostly don't ask because I'm not really that interested in a future where Linux Torvalds controls the ABI for commercial applications for my platform of choice. I don't think he takes a sufficiently long view of things, architecturally (I frankly have a hard enough time with some of the short term thinking that's happened in FreeBSD). I'd rather pursue a destiny where bad (=shortsighted) architecture isn't a fait accompli. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-advocacy" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903260234.TAA01443>