Denial is powerful thing...
Posted Feb 23, 2012 16:08 UTC (Thu) by khim
In reply to: Denial is powerful thing...
Parent article: Changes and complaints
I run rawhide at home. Sometimes versions are pushed to rawhide hours before they get pushed upstream (because the rawhide packager is also the upstream main developer, and the rawhide process is faster than the upstream process).
That changes zip in the way problem reports are treated.
Of course. Why should it change anything? Now you have unstable, constantly moving system which includes god knows what.
– “I don't run a devel stack myself, please use old stable versions for everything but my own code and reproduce” (if you can't be bothered to run the early code of others, why do you insist users should run yours before you deign read their reports)
– “your software version is too new, we'll wait to see if there is still a problem later in stabler versions”
IOW: developers sensibly expect to see how their work behaves in isolation. This is how alpha/beta software was (and is!) always tested everywhere except in Linux world. And you are most definitely are not forced to participate in beta testing, but then, when the product is released, you are stuck with decision fixable at beta stage.
Sure, integration testing is important, but this is separate issue. It's not a replacement for plain old testing.
– “please retest or I close this report” months later (by someone who clearly never bothered investigating the first report)
This, again, is separate issue. This is the question of what is better: CADT model of handling bugs or bugs open and forgotten for years. I myself prefer bugs which are open for as long as it's needed to fix them (just recently bug which I've opened on GCC tracker was fixed... five years after I've filled it), but different people have different preferences.
Please climb down from your ivory tower.
Funny that users of other OSes (and there are more of them then users of all Linux Desktop goodies combined) don't think we work in ivory tower, only Linux people expect that we need to jump through 10 times more hoops to deliver software to 10 times less people.
One example: check this instruction. See the supported versions? Right: five year old and two year old compiler. Brand-new half-year old one is not even mentioned (today it's over year old and is supported... in experimental mode). The same with Mac (half-year old Leon is not even mentioned) and the same with Linux (GCC 4.6 is not supported yet. You may run into some build errors (and patches are welcome to fix them). Please see http://crbug.com/80071 before you proceed.): an aforementioned bug is actually fixed but I suspect recommendation to use older GCC will only be officially lifted after release of Precise Pangolin.
Distribution are supposed to help users with installation of software from the repo - and this process works reasonably well - but the fact that each release has it's own repo actively hurt ISVs (because there are no process for installing anything NOT in the repo) and the fact that some pieces developed by said ISVs end up included in the distribution does not change the equation much.
In the end it hurts users as well because many of them just want one or two bleeding-edge pieces - but this is basically impossible to organize in the distributions-driven world. Some software is backported in various PPAs, but this is half-hearted effort at best: there are no way to even deliver software to Ubuntu users using PPAs unless you'll build 3-4 different packages and other distributions require still more work.
As I've initially: I'm not saying that distributions are all bad, they certainly solve some real-world problems. But they also introduce some scalability problems with their "all or nothing" approach - and this hurts everyone: ISVs, developers of upstream packages and users.
In fact this is why GNOME3-like upgrades are met with such hostility: in the distribution-driven world you only have two choices:
1. to accept new interface right away when it's not yet refined enough, or
2. to reject the change - and be stuck with obsolete versions of all other programs.
In Windows world half-backed Windows ME and Windows Vista were just skipped and people went straight to Windows XP (after suitable hardware upgrade) and/or fWindows 7 - but they had access to all "latest and greatest" goodies in the meantime. In Linux we have huge flamefests instead.
You said it best yourself: it is easier to shoot the messenger than to fix problems.
to post comments)