Posted Sep 3, 2010 4:33 UTC (Fri) by Baylink (subscriber, #755)
Parent article: Quotes of the week
I'd like to take a moment, here, to throw some liquid helium on the inexcusably silly idea that every computer on the planet should always be immediately updated to the current releases of every program running on it, notably kernels.
This fairy tale is propagated by lots of people, and it's just not so.
There are lots of very good reasons to keep running the code you are running and have tested, modulo security patches, for just as long as you possibly can.
But alas, this opinion also informs architectural design decisions (like SuSE 11.3 changing wholesale to KDE4 and Plasma with no apparent way to select KDE3.5 instead, which, y'know, works, and all my users know how to *run* it). I'm told that it's *possible* to revert to K3.5 though some manual package wrangling, but I tried it, and I couldn't get it to work. Point is: I shouldn't have *had* to. I'm running 11.0 in a couple places, and if 11.3 was going to EOL K3.5, I would expect that 11.0 should have provided K4 as an option, and I didn't see that it did. And more to the point, 11.3 should still have provided K3 as an option, and I'm pretty sure it does not.
This problem is, of course, made worse by support lifecycles.
But even when things are still in support, people's opinions of how to handle older code can be troublesome.
Certainly, the decision to run older code will always be a tradeoff.
But it seems to me that decisions to make it much harder to do so are often taken more lightly than I'd like to see.