debian: woody, sid, unstable, testing?

[snow]

alright, someone told me today, that he installs (he's supposed to be fairly knowledgable) unstable on workstations and testing on servers

alright, this seems backwards to me?

I've always, always had the impression that you put your most stable on your server-- so why not woody or even unstable (assuming testing is the least stable)

maybe I am just carrying over some sort of windows retardation, but...

anyways, he also said to stick w/ the .deb's-- so I can use apt-get; which makes sense, but is there some reason I can't install from source other than that?

anyway, comments would be appreciated. thanx

Nate Snow

[Statica]

I follow rather stringent policies that has kept my servers alive well enough:
1) Use only STABLE
2) Patch STABLE with stable

Having said that, with linux 2.2 I've rarely if ever use the stable kernels .. I got much better performance with some of the ac patches. I will not use test kernels or development releases for any component unless I really need it and the stable is severely crippled.

Also, for any server installation, I will generally try to avoid all binary distribution methods .. like debs, rpms etc etc. I still prefer compiling it myself.

Now those are just my practices and my opinions, am sure everyone has their way of working... but if I were to pay money for any data that is going to reside on any server, I wouldnt go to ppl who did what you've described above.

[snow]

thanx statica,

I thought I'd lost my mind because what he said defied not only everything I'd ever heard or read, but also logic.

he's a 16 year old that maintains the servers of a realestate firm-- he's not an expert, but he knows more about linux and networking than me, so it's hard to argue.

he was really arguing about the features

which didn't make all that much sense to me if it was a server

[Statica]

Hopefully his contract is designed in a way that he isnt held liable for crashes, or there are enough people there to support the "it just happened" theory, rather than the "why did it crash" method of thinking. Its like everything else really .. if you found a copy of Longhorn because you are on a beta testing program, would you really use that on your server? While it is a testament to Linux betas that are stable, there must be a distinction between a production system vs. a test server.