On Mon, 2003-09-22 at 00:50, Anderson, Simon wrote:
As for building from source being 'all the rage' I have to disagree. I was using ROCK some three years ago, and I believe that Slackware was compiling things from source as well long before that. It might be
I still think of package management in Linux as quite a new thing. Believe it or not, compiling from source was the only option for many years. I'm amused that there are people who assume that apt/rpm or whatever have been with us since day one.
Well, rpm has been around for a good while, and so has apt. But compiling from source, if one has a way of keeping track of what each package was creating, still has certain appeal. At the same time I can fully appreciate the argument others have that not having to compile things at all makes things easier. One could always argue that if an IT department has full control of an install server, as well as the machines on the network, they could set up an install server with packages locally compiled to match the machines on the network and have both manageability and software optimised for the hardware deployed. This is of course only really applicable to linux and *BSD where source is fully available, and things are reasonably easy to compile from source. Personally I have toyed with the idea of making my own repository of SuSE packages, recompiled for Pentium-M when gcc supports that processor fully.
however that the threshold for using a compiled-from-source distribution has dropped, making it more available to Joe User.
How the world turns.
One of the big selling points for binary distribution was that it lowered the "barriers to entry" for Joe Users as they didn't need to learn how to compile from source, could easily manage installed software and didn't have to waste time/system resource on compiles. The cost of this ease of use in terms of performance was deemed acceptable by many users/administrators as it is alleviated by continuous hardware improvements.
Very true. RPM or APT based distributions make things very easy for the system administrators whether they are home users or professionals. It also means one does not have to keep a tab on what packages just simply does not work with certain optimisation levels and which ones that has problems compiling with certain tweaks.
It struck me reading this thread that distributions like Gentoo are fashionable amongst the new breed Linux users who want to have the benefits of package management without sacrificing performance. Until reading this I assumed it was only the old time purists (particularly with a BSD background) that wouldn't ever touch pre-compiled binaries that were interested.
New, and old, linux users, perhaps more aptly described as power-users,
like to have certain control over their system. This was one of the
reasons I moved from SuSE to ROCK some years ago. SuSE did fix the
control issue, and SuSEconfig is now a bit less intrusive than what it
used to be, or it perhaps just works better. It used to drive me mad as
I would be fixing something that wasn't working, usually something
involving pcmcia (back in the day when TokenRing wasn't so easy to get
working) or things like resolv.conf, then the next time SuSEconfig ran
it would blat the changes I had made. There was not a clear-cut way of
telling it 'hands off' from what I can remember either. ROCK had the
advantage that if you didn't configure something yourself, it stayed
unconfigured, and nothing tampered with it unless you did yourself.
Perhaps this is why there is a surge in *BSD and Gentoo users. They are
confident in making things work, they do not need 'hand-holding' so much
anymore and the ports trees makes things fairly easy to compile and
accessible. To me, the main selling point of them is that you can get
stuff compiled for the system you have, rather than a generic pseudo
system aimed at running on as much as possible independent of age.
But perhaps I have a slightly different outlook on things than most
people. :)
Regards,
--
Anders Karlsson