On Tue, 2010-03-16 at 15:24 +0000, Alan Cox wrote: > > Since you have a "little bit of experience" in development :) do you > > think that developers -- maybe mainly application developers? -- would > > benefit from this deadline for downstream releases(1)? Debian's "ready > > when it's ready" developers wouldn't appreciate much, I'm afraid, but > > some agree that they must work towards more (fixed? fixe, in french) > > development periods. (I don't care much about you commenting the rest > > of my post, but I'd be interested in getting your opinion on this.) > > Again it depends. If your application wants to use cool new feature X > then you want everyone to upgrade and then use your cool new app. If you > gain nothing much from upgrades but the hassle of having to rebuild, > retest etc then 'never' is quite a good upgrade rate. > > > Hummm... I suppose every case is different. > > It's a question of benefit and timescales. It doesn't take you 18 months > to certify your desktop works and all the software on top is reliable, > make the entire set up pass a third party security audit, pass the > various credit card requirements, run performance analysis, track down > regressions and then roll out live bit by bit along with any retraining > along the way. > > Business timescales are long, and industrial timescales longer still. > There are PDP-11 systems (or these days often emulators!) still running > away in industrial plants doing what they've been doing for thirty odd > years. The machinery they are tied to is often good for fifty plus years > and depreciated accordingly, so there isn't a real urge to upgrade. > > The software folks have a very short term perspective - equalled perhaps > by only a few industries such as fashion clothing. Imagine if the first > PC you installed when you joined your first employer would be getting > decomissioned about the time you retired ? Hard to picture but in the > railroad world the chances are the first piece of track laid by some 18 > year old newbie platelayer will finally get retired about the same time > as the person who laid it. In civil engineering you often build things > that you expect to last hundreds of years. Todays engineers are doing > 'maintenance' (I guess you might consider it 'service pack 2' 8) on > victorian structures that will then be good for just light maintenance > for another century. > > This gives people a rather different sense of time and upgrading to > software engineers. > > Alan But the curves depend on the application. If one were to look at the curve for say the Wright Flyer, to the Blackbird, what would be the maturation line there? We went from 0 to 3500 knots in just about 55 years. The evolution, from the Kite like stick and cloth to stick and cloth with frameworks, to sheetmetal and rotary engines to sheetmetal and metal chassis and jets, and to composite materials and titanium with scram jets in just 75 years, and no one would think of trying to retrofit an ejection module from the blackbird to a wright flyer ;-) Whereas the genesis of the modern general purpose computer has spawned billions of devices within essentially the same time frame of the airplane (maybe a bit shorter, but one can never be sure what governments are hiding.) And while one might keep a Wright Flyer around, no one would expect it to do any serious work other than as an educational tool for budding aerospace engineers. I have programmed on PDP-11's and PDP-8's and also on microcontrollers that would run rings around them. Modern stuff gets built because there are uses and needs that the older stuff just cannot meet. Not that the older stuff has no value, just that its usefulness is not so wonderful when compared with the new kids on the block. Costs however can certainly distort this picture. Regards, Les H -- users mailing list users@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe or change subscription options: https://admin.fedoraproject.org/mailman/listinfo/users Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
