site  contact  subhomenews

Binary compatibility with a major distro

August 28, 2008 — BarryK
I'm thinking that after 4.1 and 4.2, it might be wise to move back to binary compatibility with one of the major distros. It is so incredibly time consuming compiling packages. And, you have probably noticed how many times I am recompiling the 2.6.25.x kernel. It is better to leverage off all the hard work that goes into a major distro I think.

After releasing 4.1 and shifting one step closer to retirement, I will be working on my own puplet that targets one or more of the baby laptops. I might reconstruct Unleashed and the 'devx' based on packages from a major distro -- but not necessarily Slackware, perhaps the upcoming Ubuntu 8.10 'Intrepid'. What do you think about choice of distro?


8.10 final date
Username: BarryK
Here is the release schedule for Ubuntu 8.10: Final release is anticipated for October 30. A couple of months away. That's good, I can grab it as soon as it's released and start building Unleashed from it.

Username: pizzasgood
""It is so incredibly time consuming compiling packages." This I know. I've used Gentoo. ;) And I imagine that's much more automated than Puppy + T2. What exactly does making a distro 'binary compatible' with another involve? Compiling the core libraries with the same options and versions they used? Pulling most packages straight from their repos?

binary compatibility
Username: kirk
"If you're going to base on another distro, Ubuntu seems to be the most popular, well tested and supported. I guess the only question might be which major distro builds with the least dependencies, though I suspect there's not much difference. I just rebuilt Puppy using the T2 trunk and optimized for Pentium 3. I'm using it to post this. Not easy. Had to recompile gcc, still broke when using the one T2 builds. Lots of other little problems that had to be fixed. Probably not worth the effort. Puppy will probably have to grow in size a bit if it's based on another distro, I don't have a problem with that. Loading the main sfs file into RAM (i think this is main reason to keep puppy small) often hurts performance when you need the RAM to run programs. Though I guess the first boot could be problematic. IF you retire from Puppy, development by committee will lead to a larger size anyway. I do mean "IF", puppy is pretty addictive, a bit like putting down the crack pipe :) Good luck with that! Hope your mouth is feeling better. Makes me feel fortunate about just having a crown put on a tooth last week.

Username: Southpaws
"Hey Barry, what do you think about the following distros... Sidux,Knoppix,Paldo,Debian,Mepis???

Yours are the Best!
Username: Ted Dog
"Of the many including the 'major' linux players, Puppy Linux has had the most working up to date kernel. Your kernels and Knoppix are the top of the pack. Scan the web forums and see how many other distros players refer to 'it works in puppylinux' why can't it work in ..... To embarrass the kernel workers for XYZ distro to get it working for them. Your even referenced in kernel change logs by major developers. Since your retiring, puppylinux, would you consider being a kernel master?

easy install of software? - yes
Username: Jose
"Hi everybody Does "binary compatibility" mean that a user can install software from repositories of other distribution? Will dependencies be taken care of? If yes, I think it would be a great step forward to technically less capable users. I find it quite tedious looking for missing libraries. Also my experimenting with compiling a program from scratch does not belong among my most succssfull activities. Maybe I should spend more time on it but hey - sudo apt get xxxx or sudo pacman -s xxxx is soooooo bloody easy and comfortable for everybody ;-) Or tell me if I have missed something like that in Puppy ;-) Jose

Username: lobster
"Tronkel was thinking of creating a Debian based Puppy. I must admit I never made use of slaptget - as soon as .pets were made (which was easy in 3.01 because of compatibility) I used those. MU mentioned that 4.00 (move away from compatibility) was a step back. Moving back to slackware is a move to the past. Like it or not Ubuntu is providing a Linux infrastructure. I do not hear from slackers moving to Puppy. Ubunto migrates - there have been a few . . . A few more, plus access to support, developers advice, graphic artists and the ability to influence the major player would make PuppUn2 a unifying Linux The example would be constructive and I feel sure the response would be positive. I can stop asking Mark Shuttleworth to work on Puppy. In effect we would be working together . . . Puplet for MID's? You are working on specific hardware. If Ubuntu compatible, any bloat will be offset by needing less hardware support . . . With Classmate, Eeepc and Acer hardware only . . . Simple - It becomes Puppy for the future . . . woof woof

Underpinning Puppy 5
Username: WhoDo
"If there is to be a move toward compatibility with one of the majors, I vote for Debian, which of course underpins Ubuntu. I have very good reasons for suggesting that. 1. I am suspicious of the relationship between Microsoft and the RPM-based distro's. If Novell takes the next logical step, then access to RPM-based software packages could fall off the vine. 2. OTOH, IMHO Debian-based distros remain steadfastly open and somewhat ahead of the rest when it comes to keeping up the engine-room kernel maintenance. Ubuntu is better than most because it appears to be a benevolent dictatorship, just like Puppy. I trust you Barry, but do we trust Mark Shuttleworth to remain true to the Gnu/Linux vision? 3. While Slackware remains the most "vanilla" option, the benefit of binary compatibility doesn't seem to have made that big a difference to available software packages - maybe because the Slackware package development effort appears so circumspect; not in the Puppy tradition. Leave the Lassie/Chihuahua series to maintain staid Slackware compatibility. If you do decide to hook up with an RPM-based distro, there may also be mileage in combining efforts with the PCLinuxOS-based TinyMe project. Project Leader KDulcimer already covets (in a nice way) Puppy's approach. I agree with Ted Dog's assessment that Puppy is leading the pack when it comes to keeping up with changes in the kernel for public releases. A case of fools rushing in? I don't think so; you just have less committee guff to go through to get a sensible decision. Interestingly, that may be the most significant development loss when you take that step back from leading Puppy development, Barry.

3 thoughts
Username: Bosco Bearbank
"1. Fedora development kernels tend to be bleeding edge, which makes them good for new hardware (like the Acer Aspire One I'm using to write this). Less likely to require extra modules (RIP madwifi, you served me well). 2. Ubuntu, on the other hand, tends to make make the extra modules one might require for one's kernel readily available. 3. I use both of the above distros in additiion to Puppy, and would prefer that one of them be chosen as the basis for a laptot Puplet. However, I'm sure that whatever you choose will be appropriate, and that's all that really matters

Compiling from sources is better
Username: linuxcbon
"It takes time but I prefer when everything is compiled from sources. Use the latest versions of programs if possible. 78% of used cpus are i686. 20% are x86_64. 2% are the rest (ppc, ppc64, i386). ---> compile for i686 Puppy should not rely on any other distro.

binary compatibility
Username: WN2A
"Barry, I really like the dotpet package. As long as the new Puppy(s) have stable package and good library support, then the apps can expect to have reasonable longevity. I have not seen the same with other types of packages. Regardless of how it is built, whether from source or binary, I like the small size and speed of Puppy as well. Still the favorite!

package management
Username: rockym93
"Hmmm. I quite liked the Slackware base, it opens up a lot of potential, as pretty much every software project supplies Slackware packages. With Ubuntu and Debian based systems beware that you will have to work APT in there somehow if you plan to use their .deb files... and in my experience APT is an unholy archaic mess and a pain in the butt to set up... plus getting package management systems to play nicely together is quite difficult, so you may have to abandon PET (which works quite nicely, I think) Just some stuff to think about, Best of luck with your future planning.

Package Management
Username: WN2A
"I didn't want to name names,but Barry's PET beats both .deb and .rpm , no question. That's the real heart of this distro. Tnx BK!

Best for all involved
Username: dogone
"Barry, if going with binary compatibility is easier for you, it will most certainly be easier for those who follow you. Such course corrections can only be good if they assure Puppy a long and healthy life. As to size, it's relative to the space available. The big distros are slowly finding their way onto flash devices so competition on the portable front will only stiffen. Puppy can be bigger and still offer the most bang for the byte.

Big Distro Compatibility
Username: Dougal
"If you move to basing Puppy on a big distro you should probably make it one that has a relatively short release cycle (like Ubuntu). Something like Debian just doesn't make sense, since they only have a stable release once every couple of years and you'd want to update the filesystem more frequently (and there's no point in being based on a testing/unstable branch, due to Puppy using a filesystem that's mostly "static" -- doing an update will just fill up the pup_save with newer copies of all the files in the sfs...). I'm still not sure it's the right thing to do, though. The amount of work can be cut down by delegation of authority (just like Linus doesn't actually maintain the entire kernel).

Think first, then name
Username: alienjeff
"In re: "PuppUn2" Careful in your naming conventions, kiddies. "Puppu" means "hogwash" in Finnish. Remember the Chevrolet "Nova" kludge ("doesn't go" in Spanish), and the Mitsubishi Pajero kludge ("wanker" in Spanish).

Username: Greg
"Barry: I think I am a bit older than you, and am semi retired for quite a few years.. eight years back I slacked off from four projects at one time... Puppy is small and effective... please keep it that way. One solution, starting NOW, hand off some of the grunt work... I am sure more that one person is capable of compiling and setting up the mentioned the other day having to take the time to compile different versions for testing.... Anyone interested in helping? Your expertise is primarily in research and assembly of the necessary components... your time is better spent in that area...anyone can change type face, or backgrounds or make certain icons do not overlay. Others can trace why this or that module is not loaded. And once the solution is found.. others can make the changes. I admit delegation is hard! Now is the time to start. Besides... with your tooth the whole world is not right! Hope your mouth is feeling better.. just had one pulled myself. You have started and maintained a very worthwhile project that is fun to use. It works. Step back a bit and smell the roses. You have a lot of other projects you would like to work on, the windmill for instance.... take a rest...or at least let others help! PS: In A7 Sound only runs after wizard, Zd112rw module loaded, will not see nodes, therefore will NOT connect, any help?

Username: MU
"I vote for Slackware, because the packagemanagement is well balanced, and you do not get so easy in a dependency hell as with RPM/Deb. The big distros are *based* on the big desktops (usually Gnome), what quickly can up in lots of dependencies. In Slackware, the big desktops are just optional addons. Nice to have, but I dont want to rely on them. Most small Live distros are based on slackware meanwhile, I think this has reasons :) Like When I browsed their FTP-archives, i really like how "clean" they are, everything is well separated. I really fear Gnome, as the main developers work for RedHat and Novell, and try to port Microsoft technology to the Desktop. Even small pictureviewers are written in Mono (Microsoft .net). We should not use a distro based on this philosophy I think. Mark

My Prejudice: Real Distros Build Themselves
Username: playdayz
"All respect BarryK, but I have a prejudice, that real distros build their own programs. I have watched over the last two years or so and more and more of the distros listed on distrowatch are simply remastered versions of the main distributions. It almost seems to be that many of them are just pretending to be creatred by "developers." You are the one who knows, does it contribute to Puppy's excellence that it be compiled for itself? Why did you switch from 3.01 to 4.00 (slackware to T2)? No doubt there is an hellacious amount of time spent compiling, too much for one person to sustain I am sure--but now we are talking about sharing the workload--don't we have enough good systems people to make a managable workload for those who build the OS base--and then enough applications people to take care of compiling the programs. There might be a learning curve--but that's why I am here--to learn. I have condfidence that you thought seriously about the transition to T2 and picked the best system you knew of. If those reasons are valid I would like to see at least one aspet of puppy continue to develop along those lines--as I said I believe the workload could be shared manageably.

Package Management,etc
Username: WN2A
"Mr. MU speaks true. This is sound policy. 73's

binary compatibility
Username: kirk
"I just wanted to clarify something. When I posted "Probably not worth the effort" I was referring to optimizing for Pentium 3, not the T2 build. I haven't noticed much difference in performance, but I'm also using lzma compression. But now I've got the latest base packages and Xorg. If this was something we had to do once a year or so, that wouldn't be too bad. Also, for those who might have the impression that basing puppy on another distro will mean that the other distro's packages will work in puppy, probably not.

Username: Raffy
"Puppy can identify more with Slackware than the *buntu (historically, that is). Some people here have noted that dotpet works enough for them. Then we are already able to use OpenOffice and KOffice with Puppy. We also have regulars who regularly compile software for Puppy, like muggins, etal. So either way (package compatibility or in-house compiling) should work. But am more concerned about applying the concept of "package pit" to Puppy (add packages endlessly into a pit). IMHO, Puppy does not work this way. The user selects what s/he needs, adds it, then rebuilds Puppy for personal, long-term use. Or s/he can just select what s/he needs from the ready puplets.

Package Management
Username: John Biles
"The beauty of Barry's base is that deb's, rpm's and tgz's will run fine all mixed together. TEENpup is a great example of this. For example Digikam taken from Debian 3.1, Scribus taken from Mandriva 2007, K3B taken from Slax and the list goes on. The most trouble I have is will Ubuntu packages which don't work as easily as a debian version of the same package. As for Slackware, most distro's built on it appear to run faster than their deb and rpm counterparts. But having said that distro's like debian / ubuntu have a far larger package base giving more choices to their users than slackware. The reality for me is that as long as the kernel and base libs like libc are kept up to date, I'll be able to continue using packages meant for other distro's for years to come. Barry, I think Puppy should be based on the distro that offers the most benefit in Packages and usabilty for the average PC user. While I'm not a big fan of Ubuntu, Application creators home pages normally supply a Ubuntu binary download of their app. Good Luck what ever you decide.

My opinion
Username: pizzasgood
"My personal preference would be for Puppy to be compiled from source, but T2 is something I haven't gotten around to fiddling with yet, so I don't know how much work this would involve. But like Kirk said, it's not like we'd be doing it for every release. To me the best aspect of this, aside from a stronger identity, is that we won't be bound to the choices of any particular distro. We can basically do whatever we want, without worrying if it will break compatibility with anything but our old packages (which probably shouldn't be much of a problem other than with drivers, as the important libraries are generally backwards compatible anyways). One thing people should probably notice is that this isn't a decision that Barry is making for the future of Puppy. It's something he's considering recommending, but at that point we'll be in community mode. We are the ones who will have to make the decision. I think I'll make building Puppy through T2 one of my goals for this half of the year, so that I know exactly what that will entail, and will be able to cast my vote more knowledgeably.

Puppy as a Project
Username: Viking Sailor
"The problems I see with basing Puppy on one of the large distro's are: 1) If the large distro that Puppy is based upon should change course or go under Puppy could be left on in the cold. 2) All of the large distro's already have Puppy like spin-offs. On the other hand, if Puppy is turned in to a project then it could continue to be an unique distro. Project organization might look something like this: Boss for life - Berry K. - Provides overall direction. Development manger - ???? - Works with the developers to provide new/needed functional. Package manager - ???? - Selects which packages to include in the build that provides the best functionality with the least amount of pain. Build manager - ???? - Compiles the code and works with the developers and packagers to resolve build problems. QA manager - ???? - Maintains a list of testers and works with the tester to insure good hardware and software test coverage. Paul p.s.: Sure hope you are feeling much better.

Please no bureaucracy
Username: linuxcbon
"Puppy is already a project. Please no bureaucracy : Puppy doesn't need managers but workers. In many companies, 3 workers, 10 managers, 20 secretaries. No way !

Package Compilation
Username: growler
"Maybe its the grumpiness brought on by the teeth problems that has led to all this miserable "retirement" talk! Personally I'm hoping its all a bad dream as a grateful beneficiary of the extraordinary knowledge and commitment you have bestowed on your puppy. From the various lists of people and the support shown by those people to the forum - it looks as though there is an army of people who have the ability to compile programs for puppy given the working development.sfs and a barebones puppy. Perhaps given the list of applications to be compiled these could be allocated to individuals as their responsibility to compile and uploaded to the repo. by those folks, thus taking a lot of leg work out of puppy distro building from the master and keeping him interested in the core development and the kernel "goody bag". You decide which applications and give the job to the team of supporters. I wonder if you asked for a show of hands how many would be willing to dig in to support a community effort. I hope there would be more than a few who would be willing to share responsibility. I have been building a lamp.sfs since 3.01 and will continue as long as puppy does.

Pup the best
Username: downsouth
"My ideal linux distro is small, live, installable, customisable and easily remasterable. Only PCLinuxOS mini-me versions come close to Puppy in this - and they have memory leaks. Puppy followed Vector (slackware base) for quite a while - was good, as Vector is a good, less flexible distro. But I do understand the need for package alignment.

why change what works
Username: jimpuff
"A mate and I restore old computers for resale for about $50 to raise money for those with disabilities. After lots of thought we load dingo to this vast array of makes, models and types of computers. Dingo works first up, and senses hardware accurately, 99% of the time. No other distro came close in our trials. Obviously the structure as is works, so maybe it's a case of sharing the workload a bit more, so Barry can keep on providing the inspirational spark, and overall direction

Pkg manager may be the real issue
Username: NewNameForcedByCAPTCHA
"I've just spent the last six hours trying to get expect (and, of course, tcl) running on 4.0. Although I love the ease of pup-gets when they exist, the selection is really pretty thin. In this case, Puppy's incompatibility with other distros binaries (and correspondingly, theit package managers, too) is a real PITA. Tomorrow, I'll install Ubuntu just so I can have a machine with expect installed - on Ubuntu this only requires "apt-get install expect", and magic happens! I don't have to know or care about where to get the right code for the version I'm running, or worry about dependencies like tcl. Apt-get just figures it out and 20 seconds later, it works!

Ubuntu sans politics
Username: dogone
"I continue to be impressed by the direction Linux Mint is headed. Based on Ubuntu, Mint adds a number of very user friendly and effective tools such as MintInstall and MintUpdate. If Ubuntu "binary compatibility" appeals, I urge Barry to consider achieving it via Mint. Mint offers all that is Ubuntu and more, without the paparazzi. Clement Lefebvre is an affable, progressive chap with a loyal following. A Mint partnership would also avoid the perception of Puppy as yet another Ubuntu love child.

Tags: retirement