[lug] glibc and yum update woes

D. Stimits stimits at comcast.net
Mon Aug 29 15:16:00 MDT 2005


David L. Anselmi wrote:
> D. Stimits wrote:
> [...]
> 
>> long as you don't get the devel stuff. I have no money for SuSE, 
>> debian would be tempting if not for its slow adoption of things that 
>> really are stable, mandrake is just a variation of redhat for the 
>> desktop, but I'd pay money for a properly boxed set.
> 
> 
> I keep running into people saying Debian is too slow to release.  Before 
> you drink the kool-aid, let me share my experience.

I could say the same thing about Fedora :P

> I'd have to agree that stable gets way too out of date before it 
> releases again.  So I run testing.  I upgrade to the latest packages 
> about once a month (or sooner if I want to install something new).  I've 
> always had things that are as current (sometimes more so) as Fedora's 
> latest.  I don't have much instability or breakage when I upgrade.
> 
> By default, a package spends 10 days in unstable before it automatically 

In particular I was looking at the use of 2.6 kernels way back when, and 
how long it took. I couldn't name all the packages, but over the years 
there have been a number of debian distros that would have had to be 
bleeding edge to use even remotely part of what I wanted. The main 
reason I was experimenting with FC4 and getting ready to abandon FC2 is 
because FC3 did not even have the next major release of MySQL. I came to 
find many rpm's for things I wanted to build and use, but none were 
available for FC2...and FC3 even had only part of those. I have doubts 
that debian stable would have more recent software than FC4. But that is 
only part of the reason I use the fedora/redhat derivatives.
...
> You really should give Debian testing a try.  Based on what you've said 
> I think you'll like it much more than you think.  And "etch" is a much 
> cooler name than "core 4" ;-)

If I can ever afford to get a new machine, I probably will try it. I'm 
not going to overwrite what I have now just to test though.

> 
>> Gentoo is fast, but I want to develop code to distribute, so I need
>> popular distros, and gentoo still is nowhere near as popular as
>> fedora/suse/debian/mandriva.
> 
> 
> Why does your distro matter?  If you write useful code, it will get 
> used.  If you manage to make it portable (and even more, autoconf'able) 
> the packagers who put it on their distros will even say nice things 
> about it while they use it.

I would like that to be true, but from my standpoint it isn't even 
close. I'll explain a bit below.

> Maybe I'm naive, but have you ever heard of BIND?  Runs great on Linux. 
>  Where was it developed?  BSD (what's the B for?)  Ditto for DHCP3. Does 
> Apache run on  Linux, BSD, Windows?  You bet.  Are you going to waste 
> your time porting to Windows, or on writing cool apps and let the cygwin 
> gurus port it?
> 
> Sorry, I just don't get why you think packaging past ./configure && make 
> install is useful for an application developer.  Seems like you should 
> use whatever helps you isolate your dependencies best so you can develop 
> a repeatable build process.

Very little of what I write is done in a vacuum. Often I'm helping 
people on software projects that are not mine, and which don't use 
autoconf (thankfully in some cases). None of them do anything so useful 
as bind or apache that simply having it work will get people to use it. 
On linux, there are certainly httpd substitutes, but if you eliminate 
apache, your choices for serious work are almost all gone, unless you do 
Java under tomcat...which happens to be an apache project. I can't even 
begin to tell you the number of people who say they try things that look 
ok but which they don't need only if it is in the right package format. 
Consider also the number of tarballs that become official parts of 
distros like fedora, debian, mandriva, suse, or any other number of 
distros...not a single one of them will put plain tarballs in their 
official package list and use it in their distro as is. Look at the 
admins out there, and tell me how many would abandon linux in a large 
environment if they had to manually ./configure apache and all those 
packages on a daily basis...pretty much everyone would abandon that, 
they don't have time for it.

There is no such thing as a project that can ignore marketing. Some 
projects and packages have more of their own unique ability to solve 
problems and thus force users to work harder, but I'm not working on any 
of those projects. I'm not working on anything that more than a small 
handful of people of people would bother to go through the tarball 
install (and trust me some of those are complicated config and install). 
On a good chunk of those autoconf is not even there or available. Simply 
coding and not considering the marketing of getting people to try the 
package in the first place is just another way of saying "let's waste 
some time on something that'll be abandoned."

D. Stimits, stimits AT comcast DOT net



More information about the LUG mailing list