>> >> > with the way mdk is currently working they won't be able to
>> >> > handle it.
>> >> Who said we were going to? I suppose ia64 and x64 are likely,
>> With the way things are going now, I doubt that mdk has the manpower
>> to maintain anything more than i586 at the moment.
> I think you've got it pretty much on the nose, so I don't see why this
> conversation is taking place, unless we're looking longterm future.
So, when is the right time to talk about the long term future?
> We've got PPC and x86 in the works for 9.1... I sincerely doubt
> anything else will come of that. 9.2? Who knows. I would suspect some
> 64bit port, whether x86-64 or ia64, I suppose remains to be seen.
>> >> but alpha and sparc? Those I sincerely doubt.
>> That's why it was typed between "(" ")" and a question mark was added.
>> With a little fantasy I could add a list of other archs that could be
>> done, but mdk had already brought out alpha and sparc port before.
>> Alpha is still being maintained a bit by me --> just to prove that it
>> _can_ be done --> one you have a kernel, c library, and the compiler
>> for an arch you can (re-)build the distro on it, automated.
> Well sure. I'm not saying it's a difficult thing to do, but you're
> implying official Mandrake development here.
What do you call "official"?
Redhat still compiles alpha (and sparc) in the rawhide distro (== redhat's
cooker). These packages are available. But they don't provide official
support alpha since 6.2, and sparc even longer.
Keeping a port alive (community driven or not) is different from providing
official support. Mdk can probably only support i586 at the moment, an
reading Ben Reser's article last week, I'm wondering how well mdk is able
to actually excecute.
> Alpha and sparc have typically been community-driven.. we've never had
> an official alpha port (IIRC), and sparc was last done for 7.0. Asking
> Mandrake to officially support more archs is one thing... nevermind the
> actual development cost, what about support life? If we committed to
> putting out 9.2 on x86, PPC, ia64, x86-64, sparc, and alpha, that's 5
> ports (with x86 as "main") to develop... and 5 ports to maintain for
> the life of the product.
My theory is that buildin a port doesn't have to be very expensive, at
least, if you make the investment to keep your RPMs in good working order.
The complete build process can be automated. Thus, building an extra arch
should cost close to nothing. Officially supporting it is a totally
different ballgame.
> That's a *lot*. And expensive. And for what gain? I could see it for
> a "Mandrake Server" OS, but we're typically used as a
> desktop/workstation OS. To absorb the costs of alpha/sparc/*64, which
> won't be used by many for the desktop, is cost prohibitive.
> Now, that isn't to say something can't be done to make a
> community-driven port, completely unofficial and 100% community-driven.
> Any patches to software that need to be included to accomodate another
> arch can be folded into the x86 "main" tree so that in the end the
> distrib is more portable than what we would want. But I think asking
> Mandrake to do this officially is ludicrous.
>
>> Point is, if you want to support more than one port, in a serious way,
>> then you need to start doing things differently. I think they need to
>> further automate the build process, bring in more regression testing
>> of the packages, etc. Forget uploading binaries. Only upload the
>> src.rpm's and let the backend build the binaries (on multiple
>> platforms).
>
> I understand your point. But you make the assumption that we "want to
> support more than one port, in a serious way". I don't recall anyone
> saying that this was the case. Correct me if I'm wrong, but I really
> don't believe anyone has ever said that Mandrake was officially
> interested in doing this.
Not for the alpha and sparc port, and that's why I put them between hooks.
But when x86 and ia64 come into the picture you will need to support them
in a serious way. Is mdk going to maintain those ports like the way they
doing right now with the other non-mdk ports? Or is the time ripe to
rethink the way this process is done.
> Again, I think it's a good idea, don't get me wrong. I think this could
> work very well if it is community driven. There just aren't the
> resources to do this officially and internally. Surely you must see
> that.
Yes, I see that. Mdk is more interested in wasting resources with tonnes
of manual labour, while things could & should be automated...
>> FYI: HP just brought out new alpha machines this week... EV7 based
>> CPU's, blows ia64 out of the water --> nothing new in that respect. HP
>> is trying to downplay & keep quite the performance numbers of the
>> alpha in favour of ia64.
>>
>> It's true that alpha and sparc are used as workstation and servers,
>> not as consumer desktops.
>
> Right. Do you have some comparison numbers? Ie. compare the typical
> x86 machine to an entry-level ia64 or x86-64. Then compare that to a
> similar Alpha or Sparc machine.
>
> The cost of those machines is prohibitive as well...
And yet enterprises buy tonnes of them. Keeping them running costs even
more, esspecially due to the nature of the commercial unices running on
them.
> people who can
> afford it are not using them for desktops or workstations, unless
> they're in the movie business. =)
I'd really love to run mdk on servers. Because with urpmi it's so easy to
maintain the software configuration. Why is mdk so desktop focussed? Is
there so much $$$ to be made there?
> Most people will be using PPC or x86
> for their desktop. Those are the two archs that should be focused on.
> At least both those archs are not cost-prohibitive to use as
> desktop machines.
OK.. Just wondering where mdk wants to get their $$$ from anyway. Which
market is mdk actually targetting? Consumers or enterprise users?
IMHO consumers are not going to pay for software, unless it's (M$-) taxed
when you buy the machine. The money that mdk is receiving from their
begging campains (I'm sorry, but it's starting to bug me, get a _real_
business model, please) is not structural. I for one am _not_ going to pay
for software I can download from the net. I am helping out with mdk by
uploading fixes once in a while (I think / hope I'm helping).
The enterprise, that's where the real money is. I work as a consultant for
mainly multinational companies, and don't see that mandrake is in the
picture at all. At the customer I'm working for at the moment RedHat is
the "prefered" choice for linux distro's, although no official statements
have been made about linux related product choices. When I ask about
Mandrake they see Mandrake as a niche player, with little importance.
Unfortunatly, that is. We're working with solaris there, and when I see
how much time & effort is put into simple things like software
configuration management (things that work ot-of-the-box with Mandrake) I
can't imaging how much $$$ the customer could save if he switched to
Linux. Helping them get there is where mdk can make $$$.
to summarize:
- alpha and sparc ports will not bring $$$ to mdk;
- if your RPM's are in order, adding an other port (with the right
automated rebuilding scripts) is possible, and doesn't cost much;
- if you invest in automating the rebuilding for the ports, maintaining
the ports will require less effort and more energy can be focussed on
things that can't be done in an automated manner;
- the enterprise, that's where the $$$ is;
regards,
Stefan van der Eijk
