JP: I didn't said that. In fact, I think the track Microsoft wants to follow is more or less the correct one. But they just have taken too much shortcuts and stopped too many times for a coffee and a donut...
5. Then what do you mean, when you said Linux could also be severely improved but in a different direction?
JP: The problems with Linux, at least most of them, are not directly Linux problems, but rather *nix world problems. The only big problem I see directly related to Linux, is the (un)modularity of the kernel. Linux is a monolithic kernel. In my opinion that sucks. I admit, the performance increases are big, but the functionality and the easyness of use decrease. Most of the times, when we're to add a new piece of hardware to the system, there we go to a kernel recompile. For the hardware the kernel doesn't support, some companies do provide drivers. But they need to have a version of the module for each new version of the kernel. Hardware makers don't like having too much work with drivers. And they also don't like to have the sourcecode of the drivers shown to everybody. Some companies were clever, they use the "big module technique". They write a big module with almost all of the driver code inside, with no relations to the kernel version. Then they pack it wit! h a Makefile, an install script, a few headers and .C sources to build the final module. The real driver code isn't shown. But then it's the user that doesn't like it. Sometimes it's too much work. I don't mean only for the unexperienced user. I mean for the experienced user also, because the experienced user doesn't like to be reinstalling everyting and having to resolve all the compile time dependencies everytime he's installing new hardware. Do you think sysadmins like having that much work? They like sitting down with a cup of coffee in one hand and a modelling magazine in the other, waiting for a disaster to come. When it happens, they reboot, sit down continue enjoying the coffee and the magazine.
6. That was mean.
JP: I can only hurt the ones that let themselves get hurt. Anyway, that is the reason companies do write drivers for Windows, but forget Linux. In Windows they do not have to show the sourcecode, and they only have to update the drivers on each new version of Windows, about every two years. In Linux, sometimes they would have to update them weekly. Or release the drivers using the "big module technique". It already went beyond the case of "OS not known enough".
7. And what are those things wrong with the *nix world?
JP: First of all, let me explain what I understand with *nix world. It's the applicational and interface layer of the Unixes out there. Everything above the kernel. Well, I followed a discussion a few weeks back in OSNews, about X being the biggest reason why Linux wasn't ready for the desktop. I believe that in part, that is right. X is too old. I mean X in the general way, X11, XFree86... It really sucks using X. We already do not need to use the remote connectivity X provided. IMHO it became heavy, big, with too much redundancies. I believe the next generation should connect directly to hardware, rather then going through the Client-Server layers and the exports. If we need a remote desktop... we use VNC which is small and fast. That's one thing Microsoft did right. The Remote Desktop Connection is a useful tool. But as I said, X is only one of the reasons. There are more. For example, there are too many window managers. There are two major toolkits, GTK and QT and two! major desktops, Gnome and KDE. I recently read an article, saying that Gnome and KDE started a joint effort to 'unify' the desktop manager, not join both projects, but make them use some common standards. If true, I see this as a step forward. Still not enough, but a big step anyway.
8. But you cannot deny that are *nix world problems that arise only in Linux.
JP: That's because Linux is now one of the most used *nix kernels. And because is the Linux kernel people mostly tries to use to build a desktop. And specifically about Linux, the big problem is the 'flood' of distributions that arose. I used to think Redhat was a good distro... back in the time of Redhat 5.2. It was simple, and stable. I used to like Mandrake... back in the time of Mandrake 7.1 and 8.0. Mandrake 8.1 was already bad. In fact I belive it was the worst version of Mandrake. Now, all those distros that used to be easy, aren't anymore. The 'distromakers' hunger creating a distro for the home user (which is a milestone I believe they are not achiving), but doing so they are making them hard to use by an experienced user... Why on Earth would I want to have 1001 programs to do the same thing installed on my computer? The distromakers should choose one or two, and stick to them. When installing the distro, I'll be deciding if I want any of them or not. I do not l! ike to be flooded with decisions I do not want to make and that I do not need to make. That's why I use Rootlinux, simpler and stabler than the most, only the packages that are really needed. If I want anything else, I'll go get it. And most of the work is done by hand, configuring it myself. That way I can learn more and refine my system to my personal use.
- "Joćo Paredes Interview, Part I"
- "Joćo Paredes Interview, Part II"
- "Joćo Paredes Interview, Part III"
- "Joćo Paredes Interview, Part IV"
