Tying servers and desktops together (S5E9)

[Aq]

Should distro vendors make their server and desktop products tie more closely together? Is it lock-in if all the software is open-source? What would you like to see work more smoothly? Would you go and get a specific company's server software for your home network if it hooked in more easily to your desktops?

[t0m]

I don't think that distributions should shy away from innovation, they have the resources to implement new ideas being one of the few organizations to employ staff dedicated to Linux development. I also believe that it can never be considered damaging to the community as long as the following things happen:

* obviously, releasing under open source licenses (and patent free)
* document things fully.
* reusing as much as possible: taking Aq's example using Zeroconf rather than implementing a proprietary protocol (though is less of an issue as either the new protocol will be better and supersede the original standard or there will be an eventual shift to open standards as the idea's life cycle proceeds ).

Competition within open source community is everywhere and should be encouraged as it causes projects to advance (e.g. Gnome vs KDE). The unrestricted and low barriers to entry of the ''market'' make it very easy for new entrants to overthrow the competition, just look at the rise of Ubuntu. Software is usually judged on merit and anyone can contribute resulting in a life cycle for new technologies and ideas in open source software development:

1. A first go - someone writes something cool, but its a little buggy.
2. Cloning/competition - people come along, see the idea and experiment with it, changing it and improving it.
3. Reduction - the number of projects start die away, only the ''fittest'' (for whatever metric you're using) survive.
4. Integration - the projects that survive become mature and become iconic (firefox) or become integrated into larger projects (Gnome/KDE).

Whatever innovations a distribution comes up with, if they have merit and has been released "fairly", they will find their way into other distributions and eventually become standards that everyone can work with.

[chrisp]

The original idea for this segment was should distro vendors make their server and desktop products tie more closely together at the expense of other distros and its that last bit that I'm dead set against.

If Ubuntu (say) decided to set up their desktop distro to use LDAP by default for authenticaton, and their server to serve LDAP by default, with the best admin tools bundled in then I have no problem with that, (in fact if you're reading this Shuttleworth do it!) because it doesn't lock out fedora, novell etc, it would be quite easy for them to setup their servers and desktops to integrate with that and we all win. If they designed their own "next generation" ldap equivalent then that the others couldn't include without a lot of hard work then thats a really bad thing.

Thats the microsoft way, Active Directory is LDAP and Kerberos (two open source technologies) that have been "embraced and extended" to lock out other vendors, and if you have windows desktops, you probably end up with AD, and if you've got that then Exchange, SQLServer and co make sense etc etc. And thats wrong wether MS do it, Canonical do it, Redhat do it, Novell do it, etc etc.

Open Source is about doing the right thing and playing nicely, and vendor lock in just isnt nice.

(posted from my new eeeeeeeeeeepc )

[iwuzlost]

Should distro vendors make their server and desktop products tie more closely together? Is it lock-in if all the software is open-source? What would you like to see work more smoothly? Would you go and get a specific company's server software for your home network if it hooked in more easily to your desktops?

All Linux distros should tie more closely together. WE, who use Linux, win by banning together as one and assisting each other in our advances in the latest technologies.

Lock-in is lock-in, no matter if it is open-source or not. Linux will survive and prosper, so long as we assist each other.

I wouid most definately not go with a community server's software if it hooked in more easily to my desktop. I installed and use Linux for the intellectual stimulation; not to be spoon-fed by a bunch of nit-wits who try to convince me that two pieces of horse manure with icing in the middle is actually a Joe Louis (hello Microsoft!)

That's my two cents worth.

[mrben]

All Linux distros should tie more closely together. WE, who use Linux, win by banning together as one and assisting each other in our advances in the latest technologies.

I'm not sure I 100% agree with you here, as what you describe is only a stones throw away from a virtual monopoly. We need a bit of competition there. I agree that there needs to be a certain amount of collaboration (fighting the common enemy) but if we honestly believe that Linux can "win" the desktop and/or server market, then we need to ensure that there is healthy competition there to ensure that there is continuing evolution and innovation.

Where it becomes tricky is in the current situation, where, because of the relatively small Linux market, any competition is often regarded as a "splintering" by the press. But, as stated in the show, as long as distributions continue to work with open standards and produce open source software (and, importantly, have an open development process rather than suddenly releasing an internally developed "finished" product on distribution day without telling anyone) then we will all win.

Bear in mind also that, particularly in big businesses, turnaround on server and desktops is not even in the 18 month cycle - it's more like 3-5 years. So if Ubuntu release one thing that doesn't make it into Fedora for 6 months, it's not really a big deal.

As a wise man once said, "We're not about stealing sheep, we're about growing greener grass" - we're not (or we shouldn't be) creating technologies to "steal" users from other distributions or OSes, we should be aiming to make the best technology out there, so that people can't resist the change.

[Aq]

Should distro vendors make their server and desktop products tie more closely together? Is it lock-in if all the software is open-source? What would you like to see work more smoothly? Would you go and get a specific company's server software for your home network if it hooked in more easily to your desktops?

All Linux distros should tie more closely together. WE, who use Linux, win by banning together as one and assisting each other in our advances in the latest technologies.

That's clearly a good goal, and we all agreed with that in the segment, I think. The question is this: if Red Hat invent some useful bit of software with client and server components, should they put the client on Fedora and the server on RHEL, and then let everyone else do the same if they choose to, or would you suggest that Red Hat should intentionally not do that until all Linux distros can do it at once? Doing the latter means that writing a cool useful feature is largely pointless because then you have to do twelve months of politics to get it used. Doing the former looks and feels like lock-in, even though it probably isn't because the software is open and free for anyone to implement.

I wouid most definately not go with a community server's software if it hooked in more easily to my desktop. I installed and use Linux for the intellectual stimulation; not to be spoon-fed by a bunch of nit-wits who try to convince me that two pieces of horse manure with icing in the middle is actually a Joe Louis (hello Microsoft!)

Bear with me, because I might have misunderstood this. Do you mean that you actively think that your software shouldn't get any easier to use because you want the intellectual challenge?

[fac]

That's clearly a good goal, and we all agreed with that in the segment, I think. The question is this: if Red Hat invent some useful bit of software with client and server components, should they put the client on Fedora and the server on RHEL, and then let everyone else do the same if they choose to, or would you suggest that Red Hat should intentionally not do that until all Linux distros can do it at once? Doing the latter means that writing a cool useful feature is largely pointless because then you have to do twelve months of politics to get it used. Doing the former looks and feels like lock-in, even though it probably isn't because the software is open and free for anyone to implement.

I cannot see how the former can be a lock-in. That is clearly just a matter of healthy competition. If the piece of software that Fedora has invented really is good, then Ubuntu will have it really quick.

Lock-in only happens if you have choose to make small changes to e.g. communications protocols that is not documented, or screw up the communication to other OSs on purpose. So i guess the key word here is open standards, and an open source implementation that could help others to implement the standard.

[chrisp]

if Red Hat invent some useful bit of software with client and server components, should they put the client on Fedora and the server on RHEL, and then let everyone else do the same if they choose to, or would you suggest that Red Hat should intentionally not do that until all Linux distros can do it at once? Doing the latter means that writing a cool useful feature is largely pointless because then you have to do twelve months of politics to get it used. Doing the former looks and feels like lock-in, even though it probably isn't because the software is open and free for anyone to implement.

I agree with fac, thats not lockin, thats chosing which pacakges to include with your distro, lockin is when you alter something in ways that breaks other peoples software.

So if Redhat produced a wizzy bit of software that ubuntu could use if they chose then thats cool, if they produce a wizzy bit of desktop software the prevents fedora talking to ubuntu servers without canonical ditching upstart and converting to rpms, that would be vendor lockin and a bad thing even though everything would still be open source.

[lofty]

I don't see a problem with Redhat or anyone else developing a useful bit of software and making it available for both server and desktop, if it is genuinely useful and innovative. After all the internet was originally developed on a basis of 'rough consensus and running code', not standards bodies. The issue comes where companies are behaving in a way which is clearly designed to keep them in control and push other organisations out, for example by making small changes to existing protocols, or by changing protocols every so often for no other reason than to make it hard for others to keep up. So there isn't a hard and fast line to be drawn, but you can still see from the way a company is behaving what their real intentions are.

[philder]

Do we already have this situation (of sorts) with package management? Each base distro has their own approach (Apt, Yum, Pacman, Conary* etc), so could you argue that's likewise a sort of lock in? Not saying it is, more just playing Devil's Adocate, but it's always struck me as odd there are so many potentially conflicting ways of doing essentially the same thing. If one does it better than the others, then for whatever reason is inhibited from being used globally, isn't that the same situation as the desktop/server scenario, or am I talking out of my arse?

*which I still don't even remotely understand

[iwuzlost]

Bear with me, because I might have misunderstood this. Do you mean that you actively think that your software shouldn't get any easier to use because you want the intellectual challenge?

Aq, I should have conveyed myself better than I did. I don't mind software getting easier, so long as it doesn't compromise the security and stability of my system. When this topic came up in the podcast, all I could think about was Windows and the security holes it has present in it. It also got me questioning if Linux does get easier to use, will it meet that goal by shortcuts as in Windows, or will it be done in a thorough process through the community? I truly do not want Linux to end up the same way as Windows has recently. That's the reason why I left Microsoft products behind. Sorry if I mislead you.

[Aq]

Do we already have this situation (of sorts) with package management? Each base distro has their own approach (Apt, Yum, Pacman, Conary* etc), so could you argue that's likewise a sort of lock in?

To some extent it is, yeah. However, standardising on one package *format* wouldn't help; a world with "Ubuntu .debs and Fedora .rpms" is no worse than a world with "Ubuntu .debs and Fedora .debs", which is what we'd have. There's already the problem with Fedora and SuSE, where you don't know what an arbitrary .rpm file is for, no? Ubuntu and Debian have the same issue. This can't be fixed without a desktop platform across all Linux systems, which is a much bigger problem; until then, standardising package formats can't help anyway, so why not let people experiment, is the way I feel about it.

[Foaf]

slightly OT: is there any existing automagical backup software like the one Aq described? It seems like something that one of the distros should be building.

[mrben]

Do we already have this situation (of sorts) with package management? Each base distro has their own approach (Apt, Yum, Pacman, Conary* etc), so could you argue that's likewise a sort of lock in?

To some extent it is, yeah. However, standardising on one package *format* wouldn't help; a world with "Ubuntu .debs and Fedora .rpms" is no worse than a world with "Ubuntu .debs and Fedora .debs", which is what we'd have. There's already the problem with Fedora and SuSE, where you don't know what an arbitrary .rpm file is for, no? Ubuntu and Debian have the same issue. This can't be fixed without a desktop platform across all Linux systems, which is a much bigger problem; until then, standardising package formats can't help anyway, so why not let people experiment, is the way I feel about it.

Without taking this too far off topic, surely it would be possible to do if you had a single repository that contained the debs (or rpms) for all the different distros, and a package management system that would look for the correct ones? AFAIUI Debian/Ubuntu have this to an extent already, with the split between versions. In theory, you could have all the Debian and Ubuntu debs in the same repository, and the end system would simply pick the correct binaries based on version name.

(As an (even further) aside, it would be interesting if you could tie this to the .install format that the Nokia tablets use, which automate the addition of new repositories too. If you could have a single install file that not only added repositories, but also picked up the version number/name of the distro and looked for the appropriate package, then you'd be laughing)

[Aq]

Without taking this too far off topic, surely it would be possible to do if you had a single repository that contained the debs (or rpms) for all the different distros, and a package management system that would look for the correct ones?

Yeah, but this doesn't require everyone to use the same package system to do it. PackageKit is an abstraction across different packaging systems, so you could use that to do this; it doesn't really matter whether the packages are the same format, or even in the same repository. The only real benefit of everyone using the same package format would theoretically be that you'd need fewer packages of a given bit of software, and that currently can't happen; given that that's the case, I don't see any benefit in unification (having mysoftware.ubuntu.deb, mysoftware.debian.deb, mysoftware.suse.deb, mysoftware.fedora.deb is no better than mysoftware.ubuntu.deb, mysoftware.debian.deb, mysoftware.suse.rpm, mysoftware.fedora.rpm, surely?)