Mailinglist Archive: opensuse-factory (1564 mails)

< Previous Next >
Re: [opensuse-factory] here an error, there an error, everywhere an error
On 11/10/2011 6:59 AM, Lars Müller wrote:
On Thu, Nov 10, 2011 at 12:47:40AM +0100, Rüdiger Meier wrote:
On Wednesday 09 November 2011, Brian K. White wrote:
It not our fault for needing to prevent these things, it's the
distros fault for being so thoughtlessly assembled that we have to go
through such contortions just to get installed.

Thoughtlessly assembled? Sucking community I have to reply. Fix it!
This is the opensuse-factory list. It's not leading, this is bleeding
edge. You don't have enough blood? Fix it by getting a blood bottle.

Either that or suse should just stop all pretense of being a suitable
OS for servers.

openSUSE is a community project. See my longer recent mail with
Message-id:<20111109170951.GE4977@xxxxxxxxxxxxxxx>

@Brian: Read this message two or three times please before you reply
fast and cause more unproductive noise on this openSUSE driving forward
list.

Interesting to see people complaining loud and with high frequency on
mailing lists but not being able or willing to feed bugzilla and
openFATE.

I am not the one who missed something here.

I asked, multiple times in other such threads:

Why should I?

There are already tons of other diy distro's out there, only they've been diy distro's all their lives and so they're better at it. I chose suse because it was not a diy distro and in fact was the best of the distro's that claimed not to be diy.

If I have to trade in the polished and really good suse ditro of old for the new do-it-yourself distro of today, then why use opensuse as my platform to do all this work on? It is _not_ as flexible a platform for doing it yourself as the others. It still has things in it that staps on your toes and makes it difficult to customize because until recently, you were really not supposed to need to customize it, it just worked, because they had a lot of engineers spending a lot of time _making_ it work, and keeping it working when all the underlying software and hardware keeps changing out from under it.

Take away all that dedicated and skilled staff and it is _impossible_ to maintain the former level of quality.

As I also said before mroe than once, we are getting the worst of both worlds lately. Maybe you don't know what that means or are unfamiliar with the expression.

It means, very often in life you have to choose one option or another, and each option has some advantages and some disadvantages.

In this case, way back when my company was still using 100% SCO Open Server (Even SCO Xenix before that, even Tandy Xenix before that), which also means that's what all our customers used since we are a software firm, I had of course been using linux and freebsd personally for a long time, and when one of the founders of SCO died and the companies culture and started changing and then broke up and the Unix part went to Caldera and all THAT nonsense started up, I had the ammo I needed to convince my partners to shift everything to linux.

There were many linux distros and I was already fully familiar with all of the big ones and many of the small ones.

One strategic decision that has to be made is, use a diy distro, maybe even maintain our own fully tailored and customized spin? Or use a canned distro?

With a diy distro:
Dis advantages:
You get a reasonable base to start from, the kernel major software will be updated and security patched. But a lot of things will have to be done by us, there will be little or no automagical stuff. If I want to have a samba server with a given share, I'll have to know how to find and edit smb.conf myself (ok bad example, samba has swat, then again, I'll have to know how to start and then log in to swat). If I want to print to a printer, I'll have to pick a spooler system to install, which means knowing enough about the different spoolers to even know which one I want, and then know how to configure it. Ok that's practically always cups today and cups also has a web configurator, but wasn't always the case and getting that cups web interface to actually work is not always easy. It's basically a lot of work to use a diy distro and I'll have to spend a lot of time developing our own configs and scripts and procedures, and keeping them all working as the underlying software changes over time. Also there will never be more than a ver few people in the company who would have the skill to fix problems or make changes to the system. Many tasks that it might be nice to be able to delegate, will have to be done by me or someone like me.

Advantages:
But, with a diy distro, I'll at least _be able_ to do all that because by definition it won't conflict with any of the distro's own mechanisms for managing all that stuff. And with a diy distro, I can have a really custom install that's exactly suited to our product.

With a canned distro,
Disadvantages:
Inflexible. If the supplied tools don't do it, then it's usually difficult to do it yourself because you have to not only figure out how to do the difficult thing itself, you then also have to figure out how to do it within the framework of the distro's tools or fiigure out how to trick them into doing what you want or give up and completely break the distro's tools and other config schemes, or really give up and live without whatever it is you were trying to accomplish.
The distro-compatible packaged software is only a small subset of all software out there, and a lot of random software out there may not only need to be self compiled, it will often need to be modified in some way to adapt to the way the distro is laid out.

Advantages:
At least with a good one, most of the system integration is handled by fancy intelligent scripts and utilities provided and constantly maintained by the distro. Tricky things like setting up the firewall, setting up the box to be a nat router and dhcp server, installing onto and booting from software raid, detecting and configuring the nic and printers and modems... all that has been made easy and relatively safe by some tool like yast. And, being a canned static target, more 3rd party commercial things will support it, and I definitely need that.


So there is a choice where you have to go one way or the other, and both ways have some advantages, and both ways have some disadvantages.

Worst of both worlds means when some path has some or all of the disadvantages of one way, AND some or all of the disadvantages of the other way, and none of the advantages of either way.

That is what I am saying has happened to opensuse.

I'm FINE with diy distros. You get what you put into them.
And in trade for that requirement that you do everything yourself, you are allowed to to things yourself.

What we HAD was a distro where you were not easily allowed to do much yourself, but in trade for those limitations, you didn't have to worry about making the basic system work most of the time. Most hardware was supported, most services had reasonably functional configurator front ends, you didn't have to know what grub even is. People back at HQ spend all day figuring out that if you have a particular chip in your nic, then you need to put a particular option in modules.conf or on the kernel command line in the bootloader, and also disable some checkboxes in some of the network manager screens because your may claim to offer some feature like offloading tcp checksums, but in reality it doesn't work well enough so you have to know not to use it...

What we HAVE now is a distro with most of the old limitations and few of the old benefits. Now when something doesn't work it's fix it yourself. When something could work but requires a change to the distro's tools to stop doing something wrong, it's figure out for yourself why it's broken, how to fix it, and then request the few remaining hq guys to please do this fix that you figured out for them, and then they never do it.

That is a bad deal.

--
bkw
--
To unsubscribe, e-mail: opensuse-factory+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse-factory+owner@xxxxxxxxxxxx

< Previous Next >