Show me the money...

Hi Joe,

Sorry for addressing it this way, but running a business is all about the money, right? With that, how much would it cost to get V/Wmin up to date for Mandriva 2006? I can appreciate that a lot of it has to do with market share, I think this might be a chicken and egg situation, with Mandrake merging/purchasing Conectiva and forming Mandriva many are reluctant to start offering support.

Please let me know your thoughts.

Best regards,

Hey Charles,

I understand your impatience, as every new OS that we add is taking much longer than I originally anticipated–mainly due to other distractions, like dealing with bugs on the existing platforms, but also due to the difficulty of doing things the right way on each platform…they’re all a little different. Mandrake branched off of Red Hat, of course, and so there are similarities, but it has been many years since that branche and much has changed as well. I am currently working on Mandrake and Debian, and there really isn’t much that can make it go faster…i.e. it’s not really a question of whether it will pay for itself, as we’ve committed to supporting it before we end EA (which is when we get to reward ourselves by charging full price and advertising to a wider audience…it’s also when I suspect we’ll actually begin making a reasonable wage doing this work, so we already have very strong monetary motivation to get it finished as fast as possible).

In short, I don’t know of any way to make development on new platforms go much faster, short of doing it the wrong way (the way some of our competitors go about it–a bunch of clunky build scripts and oddball installation locations interspersed with the occasional proper, but generic, package). Once we’re up and running on a platform and the platform dependent bugs are resolved, maintaining it is quite easy…The Red Hat based platforms have settled down quite a bit now, and aside from some issues due to me deploying broken ClamAV packages to start with that blew up last week, it’s becoming easier by the day.

Anyway, today is a porting day, so it’s likely that we’ll get much closer to a Mandrake release by this evening. I can’t promise a date (I’ve tried doing that in the past and ended up being way off every single time)…but progress is being made and I don’t expect Mandrake to be difficult. It has good package management, similar initscripts to Red Hat, and a lot of the packages we need are provided by the vendor.

Hey Charles,

You’ll be happy to know that due to a vicious sore throat, I’ve been unable to sleep tonight (that’s not the part you’ll be happy to know…I hope), and I’ve spent the time making the installer spin for Mandrake. I’m performing the first fully automated installation on the system right now.

At least a day or two will be required for testing (and I’m afraid Mandrake is proving shockingly buggy in places that tend to slow down development quite a lot–it may be due to running it under Qemu, but none of the other distros have exhibited this kind of trouble*), but even with the slowdowns I’m comfortable committing to an installer useable with Mandrake 10.2 before the week is out.

*-The bugs I’m seeing involve any network mounted filesystem. I normally mount up my local mirror of my repository on the OS and use the native OS to generate the package management index (using genhdlist on mandrake, genIS_PLAINcache on SUSE, etc.). But on Mandrake it simply locks when I do this. Obviously, this also makes it hard to build packages (since the SRPMS are in the repo on the network file system), copy built packages back into the repo, etc. It also seems to panic during boot about 20% of the time. I’ve yet to figure out what makes that one happen. Oh, yes, and shutdowns hang 100% of the time and eat up all of the CPU on the host system as well and the qemu process has to be killed manually. Quite ornery. Ah, good, now the install has stopped with errors from the Mandrake software repository–I have no clue how to deal with missing dependencies in the mirrors, not to mention there doesn’t appear to be a way to get urpmi to actually automatically resolve dependencies without getting confirmation short of using --force. (Can you tell I’m finding the whole Mandrake experience rather trying? For those who love Mandrake, just write this off as the grouchy ramblings of someone who is too sick and tired to be working, but can’t seem to get to sleep. I’m sure it’ll look far nicer and less buggy when I’m feeling better.)

Hey Charles,

In case you don’t subscribe to the news forum, or check the front page, I’ll also post here:

Mandriva 10.2 on i386 is now supported by the installer. Let me know if you need x86_64 (I’ll get around to it regardless, but if I know someone needs it I’ll get around to it faster).

Mandrive, née Mandrake.

Where do I begin…

> (and I’m afraid Mandrake is proving shockingly buggy
> in places that tend to slow down development quite a lot

This does not come as a surprise, as Mandrake has been plagued by an incredibly buggy development since I first laid eyes on it, and every couple of years, when I give it another shot, it turns out to still be plagued by a plethora of problems, and a developmnet team that seems unfamiliar with the concept of ‘mission critical’.

Mandriva is great as a desktop distribution, but if your intent is to use it for mission critical applications (like, hosting), you’re taking your chances.

Hat off to Joe for spending the extra time to make the installer work with Mandriva, but, as they say, ‘you have been warned’.