Some people seem to think that Canonicals greatest obligation to people is not to upset anyone in anyway. If you’re Ubuntu user and feel somehow betrayed because of this Mir thing, why don’t you install some other distribution and let it go.

For a long time video card support in Linux/*BSD lagged and just having reasonablly stable and performant 2D support would be considered lucky, forget any notion of 3D. Over the years, that situation improved and it got to a point where most of the 2D hardware worked reasonably well and some basic OpenGL support existed for some of the more popular 3D hardware.

Then came the turning point. New video cards are all about 3D and treat 2D as some legacy thing that might be used to boot the computer or whatever. Somebody realized that the new hardware needed to be thought of differently, but didn’t bother to consider the importancre of backwards compatibility. Suddenly I find myself in a situation where only the newest video hardware has any sort of support and it’s stillin development so it is buggy and not performant. Meanwhile the only way to have ANY OpenGL in hardware, or in saome cases ANYTHING more than VESA, is to use an old X server and old Mesa that are not maintained. That means any remaining bugs are not addressed even if patches are supplied by the users. That also means that some applications are stuck at older versions, not receiving patches that may leave them with glaring vulnerabilities.

When you write off old hardware, you forget two critical things. 1) Not everyone can afford to buy the newest every year. 2) Not everyone wants to throw away hardware that works just because it isn’t new. Point 2 is especially valid if you consider that the older stuff was built better. I have boxes that are more than a decade old that run rock solid, and I have some stuff built in the last 2 years that has some portion that doesn’t work because it either never did or some component burn up/malfunctioned in a matter of months. Between those points there is a gap as the hardware made in that time period is all dead.

Especially for video hardware, you have to consider that not everyone has the newest GPU. Many server motherboards have some rather old but “well supported” (prior to the KMS move and Mesa “reboot”) video hardware onboard and might not have much option other than a PCI slot for alternate video hardware. Sure there are plenty of people that say you shouldn’t even run X on a server, but sometimes you just have to fire it up for a task (installing certain software packages for example). On several of my servers that have “modern” onboard video, I disable it and slap in an old PCI card just to free up memory bandwidth. For example, my file server has intel video in the chipset, which uses system RAM and uses 3% of the memory bandwidth to just draw a 80×25 textmode console with no screen activity. All the PCie slots are full of network cards and disk controllers, so I put a Matrox Millenium 2 in a PCI slot. That frees memory bandwidth for the stuff that needs it (serving files), I can run a higher res console, and the video card is actually performant in text mode (the new stuff that treats 2D as legacy might nnot have working VESA, and if it does, scrolling a full screen of text can be so slow as to make a noticeable difference in kernel compilation time just due to the burden of rendering the compiler output inefficiently). Also, if you actually use X11 over a network connection as it was meant to be, all this 3D crap falls on it’s face.

There is also a disturbing trend in development where not only is it assumed that all users will constantly buy new hardware and not desire to keep using the hardware they already spent good money on, but that the technology progress will keep pace or accelrate. We have already seen CPU clock rate peak and actually back off as the manufactureres upped the core count. Not everything can be parallelized efficiently and we see plenty of evidence of that. The notion that users will have the same ultrafast box that some developer has shortly after the code is done and released is a fallacy. Even if it were true, software bloat has outpaced hardware development in the past decade.

I recently built a monster of a desktop and it still just FEELS slow on some things that should be fast. Sometimes I think I might be deluded by my memory, but all I have to do is fire up one of my 10+ year old boxes and watch it fly. I boot up my SGI Indy, running the last IRIX 6.5.x that supports it (released over a decade later than the hardware) and it flies compared to any Linux desktop, or any Mac. Even Windows 2000 on a P2 feels fast in comparison. What really blows my mind is my cell phone: 1Ghz CPU, 1GB RAM, 64GB flash, it runs Linux and it feels slow. Switching apps on it takes seconds as does launching apps. My first PC was a 486 with 24MB RAM and 2GB harddrive (after upgrades) and it ran OS/2 like a champ. I can fire up OS/2 on a P2 with 512MB RAM and some absurdly large disk and it too flies, apps launch in fractions of a second and task switching is instant. I remember when Linux was the efficient little thing you put on the old box that worldn’t run the new Windows version too well. Now it’s a monster that performs worse than anything else (except maybe the BSDs in terms of video which are largely still using UMS for X11 or VESA if you are unfortunate enough to be stuck on new hardware).

I run KDE4 on my desktop because I want some of the apps and it’s good enough. It’s still a crap interface compared to KDE3, but it has become untenable to keep KDE3 working, and many new apps won’t run on it (but there are a few good ol’ apps that don’t work on KDE4). On everything else, I run Enlightenment, ironically. Ironic because when I first installed Linux with a 2.0.x kernel on my 486, Enlightenment was the eye-candy WM that was dog slow compared things like FVWM and even KDE1 when that came around (yes, I’ve used KDE through every version). Now, Englightment is the quick one that does just fine rendering on the CPU at the maximum resolution of a 15 year old video card.

The purpose of making faster hardware was to do the same tasks as before but quicker, and to enable one to do new tasks that were previously not practical. What can I do now that I couldn’t do a decade ago? Nothing really. The difference? What I could do before I still can but slower. Wait, slower? Yes, SLOWER! I don’t need a fancied up interface that takes so much of the system resources that I am waiting on it. THAT is what is perceived as “heavy” and what we want to avoid with “lightweight” alternatives (not that any you mention are noticeably lighter).

(thanks auto-spellcheck, now where is the edit post button?)