Re: What's a developer to do?
Jon Hall wrote: >>Well, LSB is Linux-only. > > No. BSD and Solaris systems can also pass the LSB. OS X could pass it if > they wanted to. Interesting. I wasn't aware of this. > LSB simply defines a binary interface for applications to run.and it > does it on a architecture basis. > >>Starpacks will run on Linux, OS-X, Windows, ... >>- and *with the same binary*! > > U, I think you mean that the *envelope file* Starpacks creates will > deliver the binaries needed for all these platforms, if you have the binaries, > by utilizing the TCL interpreter. Correct? Technically, yes. (The end-user sees it as the "same program", and for the most part, the developer does too). > You still have to have the binaries of the application itself for a particular > OS and architecture, and in the case of Linux, it would be nice if that > application followed the LSB, and if the platforms you were delivering it for > were LSB compliant. Those "binaries", if there are any, are bundled into the application's "starpack". (Typically, each platform's TCL is the only binary in a starpack). I've been assuming the starpack's SDX utility (which glues the various parts of TCL and the application into a starpack) was LSB compliant and created LSB compliant binaries - at least for the Linux platform. That assumption may not be valid. > Of course I do not see where Starkits does any of the testing for > prerequisites > and dependencies that RPM or APT does. It doesn't need to. RPM's, APT, and other Linux packaging concepts are beyond the scope of starkits. Actually, starpacks are another form of RPMs, ... but mostly for people developing on systems that don't (or didn't) have a concept of package management. In TCL, package dependency management is done with the "package" command, and its done at run-time by the application, instead of an external packaging system. BTW: I just read on the starpack mailing list where someone suggested that jcw use RHL 7.3 as the basis of starpack development. Evidently, Linux binaries developed on a RHL 7.3 system will run on most other distributions. They went further to suggest that any application that doesn't work from there should be rebuilt from scratch for the target system. I suspect this is the direction jcw is going to take. --Bruce ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
>Well, LSB is Linux-only. No. BSD and Solaris systems can also pass the LSB. OS X could pass it if they wanted to. LSB simply defines a binary interface for applications to run.and it does it on a architecture basis. >Starpacks will run on Linux, OS-X, Windows, ... >- and *with the same binary*! U, I think you mean that the *envelope file* Starpacks creates will deliver the binaries needed for all these platforms, if you have the binaries, by utilizing the TCL interpreter. Correct? You still have to have the binaries of the application itself for a particular OS and architecture, and in the case of Linux, it would be nice if that application followed the LSB, and if the platforms you were delivering it for were LSB compliant. Of course I do not see where Starkits does any of the testing for prerequisites and dependencies that RPM or APT does. md ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
Bill Sconce wrote: > On Fri, 21 Apr 2006 13:17:07 -0400 > Bruce Dawson <[EMAIL PROTECTED]> wrote: >>There is a real, but subtle, problem here, and I thank maddog for >>seeing it and looking into it more. > > Back to serious. Yes, there is a problem, and I second the vote for > maddog's sentient take on it. I.e., we might read Jean-Claude Wippler's > concerns as an affirmation of the need to make LSB real. (It's hard > to tell through the whining -- is there a subtext of "I want starpacks > INSTEAD of LSB"? That would be very Unix-wars like itself, in a > self-referential way.) Well, LSB is Linux-only. Starpacks will run on Linux, OS-X, Windows, ... - and *with the same binary*! My DOJ contract is starpack based, so it would be nice if I don't have to dictate to the client "thou shalt run on Linux", or "thou shalt run on Windows", or ... - you get the picture. >>I shall stick to technical details on future postings to this thread. > > No, no... this is gnhlug... :) Oh yeah. Right. :-) --Bruce ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
On Fri, 21 Apr 2006 13:17:07 -0400 Bruce Dawson <[EMAIL PROTECTED]> wrote: > >>(what a total cop-out compared to Windows!). > > > > Disconnect. > > Power off. > > Bye. > > Ah Ha! Now who is doing the cop-out? Er, me, of course.:) > PS: I'm using "fanatic rhetoric" because people on this list seem to > respond to it Q.E.D. :) :) > (And I'm missing my geeky friends lately - umbrella-laced rum only > goes so far). Same here. Hurry home, Bruce. > There is a real, but subtle, problem here, and I thank maddog for > seeing it and looking into it more. Back to serious. Yes, there is a problem, and I second the vote for maddog's sentient take on it. I.e., we might read Jean-Claude Wippler's concerns as an affirmation of the need to make LSB real. (It's hard to tell through the whining -- is there a subtext of "I want starpacks INSTEAD of LSB"? That would be very Unix-wars like itself, in a self-referential way.) > I shall stick to technical details on future postings to this thread. No, no... this is gnhlug... :) -Bill ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
Bill Sconce wrote: >>"I'm starting to lose the battle with Linux - the dynamic builds will >>not work on all Linux systems, and the static builds are doing such >>nasty things in libc nowadays that they too probably won't work without >>specific libc.so's on your system. Apparently the world is moving >>towards a state where only Linux distro builders can produce proper >>binaries, > > Ah. Some weasel words to be aware of: "Apparently", "the world is moving." Ah. I saw the weasel words. The original author, Jean-Claude Wippler (jcw) is meritous for noticing a barely perceptible slippage that's happening in Linux, and to be applauded for pointing it out so succinctly. That is: Linux is diverging into the old Unix Wars mentality. As maddog pointed out, LSB should solve this. Jcw was lamenting that the problem still exists, regardless of LSB. I think maddog will be determining if starpacks/TCL/... are following LSB conventions or not. >>and where a binary no longer works across Linux releases, > Ah, he IS trying to spin us up. (Binaries NEVER "worked across releases". > (At one point we broke ALL of them, on purpose ... remember ELF?) Of course I'm/he is trying to "spin us up" ... the original posting points out a failure in the Linux culture - its in serious danger of regressing from its "shiny city on the hill" back to the dank hallways of the cathedral. The original author was lamenting that Linux appeared to be moving away from the problems of incompatibility, but now no one appears to care anymore. The move to ELF was something the community *wanted* to happen - it solved *a lot* of architecture compatibility problems. And I have to admit that I've been seeing a lot of the "I just want software, I don't care if it works or not" attitude creep out of the consumer marketplace and into the Linux space. This is part of what made Microsoft the bumbling, yet profitable giant it is today. Do we want Linux to be just like Microsoft? I believe some of us do, but I think those people can be coaxed to think twice: do we really want Linux to be like Microsoft? (i.e. do we want Linux to be a business?) >>(what a total cop-out compared to Windows!). > > Disconnect. > Power off. > Bye. Ah Ha! Now who is doing the cop-out? --Bruce PS: I'm using "fanatic rhetoric" because people on this list seem to respond to it; seems to be part of being a "social club for geeks". (And I'm missing my geeky friends lately - umbrella-laced rum only goes so far). There is a real, but subtle, problem here, and I thank maddog for seeing it and looking into it more. I shall stick to technical details on future postings to this thread. ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
Bill, Bruce did go back to the developer, who fed me more information, but while I have not had time to look at the issues in depth, the impressions that I have fed to this mailing list are the same. I do intend on going back and looking at their issues in depth, but I have to find the time to do that. md -- Jon "maddog" Hall Executive Director Linux International(R) email: [EMAIL PROTECTED] 80 Amherst St. Voice: +1.603.672.4557 Amherst, N.H. 03031-3032 U.S.A. WWW: http://www.li.org Board Member: Uniforum Association, USENIX Association (R)Linux is a registered trademark of Linus Torvalds in several countries. (R)Linux International is a registered trademark in the USA used pursuant to a license from Linux Mark Institute, authorized licensor of Linus Torvalds, owner of the Linux trademark on a worldwide basis (R)UNIX is a registered trademark of The Open Group in the USA and other countries. ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
I just noticed. The original post was two paragraphs, one by a Starkit author, one by Bruce. And reading the first one more closely I wee why maddog found the complaint "loose". > "I'm starting to lose the battle with Linux - the dynamic builds will > not work on all Linux systems, and the static builds are doing such > nasty things in libc nowadays that they too probably won't work without > specific libc.so's on your system. Apparently the world is moving > towards a state where only Linux distro builders can produce proper > binaries, Ah. Some weasel words to be aware of: "Apparently", "the world is moving." (With apology/credit to Bruce, who noticed similar dross in another thread.) > and where a binary no longer works across Linux releases, Ah, he IS trying to spin us up. (Binaries NEVER "worked across releases". (At one point we broke ALL of them, on purpose ... remember ELF?) > let alone on another distro Weasel words to be aware of: "let alone". > (what a total cop-out compared to Windows!). Disconnect. Power off. Bye. ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
[EMAIL PROTECTED] said: > If that's correct, a developer who tries to package binaries for every > distribution has chosen a difficult task. A few organizations (such as > Mozilla) can do it, but it's a big deal. > If a developer asks that all distributions synchronize libraries isn't she > wagging the dog, getting it kind of backwards? This is what LSB is supposed to make better/easier. LSB specifies that for a given level of LSB, the distribution has to use a certain level of libraries. If a distribution is "LSB compliant", then it uses that level of libraries, and you therefore know that "Red Hat VX.x" and "SuSE VY.y" both of which are LSB VZ.z compliant should be able to run the binaries from TCL VW.w The FSG studies these things enough to note where binary inconsistencies will be and to warn groups that they are coming. TCL should be building to the LSB, not to a specific distribution. If more layered applications and languages did this, it would send a very loud and clear message to the distribution vendors to pull into line behind and support the LSB effort. One of the problems is that this is complex, and the whole thing is still under development and transition. Not all distributions build and test against the LSB, and this creates issues for people like the TCL developers that want to support all distributions, not just LSB compliant distributions. Of course we could just solve this problem by people abandoning the use of any distribution other than Red Hat. Then we just rename the company to "Microsoft". md -- Jon "maddog" Hall Executive Director Linux International(R) email: [EMAIL PROTECTED] 80 Amherst St. Voice: +1.603.672.4557 Amherst, N.H. 03031-3032 U.S.A. WWW: http://www.li.org Board Member: Uniforum Association, USENIX Association (R)Linux is a registered trademark of Linus Torvalds in several countries. (R)Linux International is a registered trademark in the USA used pursuant to a license from Linux Mark Institute, authorized licensor of Linus Torvalds, owner of the Linux trademark on a worldwide basis (R)UNIX is a registered trademark of The Open Group in the USA and other countries. ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
Bill Sconce wrote: > On Thu, 20 Apr 2006 19:17:48 -0400 > Jason Stephenson <[EMAIL PROTECTED]> wrote: > >>Frankly, I think that is what the different distros are for, providing >>binary packages that work with their mix of software and libs. > > I agree. It's the flip side of a user's perennial question: e.g., > "should I wait for Debian to package XYZ for me or build it myself?". > >>They [the distros] *should* be doing the work of distributing the >>binaries. > > If that's correct, a developer who tries to package binaries for > every distribution has chosen a difficult task. A few organizations > (such as Mozilla) can do it, but it's a big deal. > > If a developer asks that all distributions synchronize libraries isn't > she wagging the dog, getting it kind of backwards? Except, if you're a developer, you want your software to run on all platforms, regardless of architecture, libraries, ... This is what Starpacks provide. But this is becoming increasingly difficult to achieve, which is what the original author is lamenting. Has the business of software development (and resulting desire to lock-in customers) become so ingrained that it has caused the industry to loose one if its Holy Grails? --Bruce ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
On Thu, 20 Apr 2006 19:17:48 -0400 Jason Stephenson <[EMAIL PROTECTED]> wrote: > Frankly, I think that is what the different distros are for, providing > binary packages that work with their mix of software and libs. I agree. It's the flip side of a user's perennial question: e.g., "should I wait for Debian to package XYZ for me or build it myself?". > They [the distros] *should* be doing the work of distributing the > binaries. If that's correct, a developer who tries to package binaries for every distribution has chosen a difficult task. A few organizations (such as Mozilla) can do it, but it's a big deal. If a developer asks that all distributions synchronize libraries isn't she wagging the dog, getting it kind of backwards? .025 -Bill ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
Uh, he could release the source and let people build it themselves. That is honestly the only way to guarantee the code runs on your machine, to compile it yourself. Frankly, I think that is what the different distros are for, providing binary packages that work with their mix of software and libs. They *should* be doing the work of distributing the binaries. I must say, though, that I've never really had an issue with getting binaries to run on GNU/Linux or FreeBSD. I've always been able to resolve the depency issues by reading the errors when a dynamically linked application fails to link. The most trouble that I recall having was with getting Java to run on a Slackware box, and that was a simple matter of soft linking a couple of binaries from where Slackware puts 'em to where Red Hat would put them, and then running ld on a couple of the Java binaries so that the link info would get updated. Anyway, I know this isn't helpful but I'm starting to agree more and more with RMS (and probably Maddog, too) the more that I deal with binary only "software." I've recently adopted the motto that it isn't software without source code, so I have little sympathy for the difficulties that developers face getting their binary-only software to work on different platforms. Sorry for the polemic. I'll step down off of the soapbox, now. I'd suggest to your developer that he actually consider running ld as part of the post-install. That just might fix a lot of the problems. Cheers, Jason ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
Re: What's a developer to do?
Jon maddog Hall wrote: > Bruce, > > I would be interested in a pointer to the author. His verbiage is a little > "loose" for anyone to understand completely: I've attached his message to this one. >>the dynamic builds will not work on all Linux systems >>Is LSB dead (again)? > > What does he mean by "all"? All LSB V3.x compliant systems, or XYZ Linux > distro that has done no testing or certification against LSB? He never mentioned LSB - I did because my understanding was that it is supposed to solve this kind of problem. > Are the libraries that he is trying to get to work outside the LSB > specification? Not unless glibc is outside LSB! But I suspect his complaint arose because someone tried to run the application on a "common" linux system and ran into library version skew. >>and where a binary no longer works across Linux releases, let alone on >>another distro > Again, what is a "Linux release"? A release of the kernel? A release of a > new LSB spec? My guess is that a "Linux release" is meant to be GNU/Linux; a combination of kernel, libraries, configuration files, and associated "glue" programs to make it operate as a cohesive unit. > Is *his* application written to LSB? The LSB plainly states that if you use > libraries and interfaces outside its realm, that you have to include those > libraries with your application. His application is TCL. Evidently, he runs into problems even if he uses static libraries. Since its designed to be platform agnostic, it needs to run on not just LSB systems, but Windows and Macintosh. He's lamenting that its easier to build and distribute an application for Macs, Windows and other platforms than it is for Linux. > I don't mean to start a diatribe on this subject, but I would appreciate > getting more information from the author. Thanks for responding, but I really wasn't expecting a response from you! I was expecting other GNHLUGers to say "what's the problem" or "But TCL is way out on the fringes, so who cares" or something like that. But I have to admit that I've run into similar problems in the past, and switched to Java in an attempt to solve them. Then I went back to TCL because it would actually run on dissimilar systems and with a much smaller footprint. > Thanks, And thank-you! --Bruce > md > === > From: Bruce Dawson <[EMAIL PROTECTED]> > Date: Thu, 20 Apr 2006 09:02:23 -0400 > To: GNHLUG > > I ran across this on the Starkit mailing list... (Starkits are a way of > packaging TCL modules so they are portable across architectures and OS.) > > The author laments: > > "I'm starting to lose the battle with Linux - the dynamic builds will > not work on all Linux systems, and the static builds are doing such > nasty things in libc nowadays that they too probably won't work without > specific libc.so's on your system. Apparently the world is moving > towards a state where only Linux distro builders can produce proper > binaries, and where a binary no longer works across Linux releases, let > alone on another distro (what a total cop-out compared to Windows!). " > > I thought this battle was fought (and won by at least VMS) back in the > 80's. What happened? Is no one fighting the upgradability challange > anymore? Isn't Perl, Java, Python, ... having similar issues? Are > individual corporate interests winning out over general user/developer > interests? Don't Linus and RMS care anymore? Is LSB dead (again)? > > --Bruce --- Begin Message --- FYI, there are new Tclkit 8.4.13 and 8.5a4 builds at: http://www.equi4.com/pub/tk/ The sources used are at: http://www.equi4.com/pub/tk/tars/ See the download matrix for a breakdown of the different binaries: http://www.equi4.com/pub/tk/downloads.html Note that some of the 8.5a4 builds are older ones, you should be able to see the file date by hovering over the entry with the mouse. Main changes are new Mac OS X x86 builds and Universal builds (combined ppc/x86 binaries). I've also added a couple of Aqua builds, but keep in mind that these are still tclsh-like, i.e. they bring up a console and get "promoted" into a gui when doing "package require Tk". There are several "lite" versions (i.e. no Mk4tcl C++ code) in the lite/ subdirectories. These are sufficient to use starkits/starpacks read-only, and SDX has some extra code to be able to wrap (i.e. *write*) starkits using Tclkit Lite (tho I've not tried it with the latest builds yet: SDX may need to be tweaked). I'm starting to lose the battle with Linux - the dynamic builds will not work on all Linux systems, and the static builds are doing such nasty things in libc nowadays that they too probably won't work without specific libc.so's on your system. Apparently the world is moving towards a state where only Linux distro builders can produce proper binaries, and where a binary no longer works across Linux
Re: What's a developer to do?
Bruce, I would be interested in a pointer to the author. His verbiage is a little "loose" for anyone to understand completely: > the dynamic builds will not work on all Linux systems > Is LSB dead (again)? What does he mean by "all"? All LSB V3.x compliant systems, or XYZ Linux distro that has done no testing or certification against LSB? Are the libraries that he is trying to get to work outside the LSB specification? > and where a binary no longer works across Linux releases, let alone on > another distro Again, what is a "Linux release"? A release of the kernel? A release of a new LSB spec? Is *his* application written to LSB? The LSB plainly states that if you use libraries and interfaces outside its realm, that you have to include those libraries with your application. I don't mean to start a diatribe on this subject, but I would appreciate getting more information from the author. Thanks, md === From: Bruce Dawson <[EMAIL PROTECTED]> Date: Thu, 20 Apr 2006 09:02:23 -0400 To: GNHLUG I ran across this on the Starkit mailing list... (Starkits are a way of packaging TCL modules so they are portable across architectures and OS.) The author laments: "I'm starting to lose the battle with Linux - the dynamic builds will not work on all Linux systems, and the static builds are doing such nasty things in libc nowadays that they too probably won't work without specific libc.so's on your system. Apparently the world is moving towards a state where only Linux distro builders can produce proper binaries, and where a binary no longer works across Linux releases, let alone on another distro (what a total cop-out compared to Windows!). " I thought this battle was fought (and won by at least VMS) back in the 80's. What happened? Is no one fighting the upgradability challange anymore? Isn't Perl, Java, Python, ... having similar issues? Are individual corporate interests winning out over general user/developer interests? Don't Linus and RMS care anymore? Is LSB dead (again)? --Bruce -- Jon "maddog" Hall Executive Director Linux International(R) email: [EMAIL PROTECTED] 80 Amherst St. Voice: +1.603.672.4557 Amherst, N.H. 03031-3032 U.S.A. WWW: http://www.li.org Board Member: Uniforum Association, USENIX Association (R)Linux is a registered trademark of Linus Torvalds in several countries. (R)Linux International is a registered trademark in the USA used pursuant to a license from Linux Mark Institute, authorized licensor of Linus Torvalds, owner of the Linux trademark on a worldwide basis (R)UNIX is a registered trademark of The Open Group in the USA and other countries. ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
