Le samedi 18 novembre 2006 à 08:37 +0200, Avi Kivity a écrit : > Bernardo Innocenti wrote: > > Nice troll, but there's a point that can be made out of it: > > > The fact that these applications originated as proprietary software is > > still very recognizable today. > > What's recognizable is that they're user-oriented instead of developer > oriented. No, what's recognizable is they're PHB-oriented, PHB asked for a single tool, PHB got delivered a single tool, even at the cost of tying vastly different pieces of code with bandaids > > They still are very monolithic in nature, > > Users like one interface to handle all their needs. Developers | want | > to | connect | orthogonal > tools. In my experience users are quite happy with several interfaces as long as they're consistent and not overlapping. It's developers that insist on making private copies and forks of uncounted libs, then slap an "unifying" GUI over the mess (in a FLOSS context they get reminded of proper practices by distributions which hate having to ship and qualify multiple versions of the same thing; in a proprietary context they're given free run). The "users like one interface" is a myth, firefox succeeded because it broke the old mozilla in multiple interfaces, and before that moz was eating the netscape-branded version which had welded yet more stuff in the single interface. > > built on their own custom portability and GUI frameworks. > > Most users don't use GTK That's why using one-of-a-kind proprietary framework is better ? One of the clearest and most appreciated changes in OO.o 2 was to expose native widgets to users instead of the old SO thing. Mac users BTW complain because they were forgotten in this change. > > They make heavy use of binary or opaque file formats for storing > > settings and other metadata. > > Users want to configure using a gui. That means a program reads and > writes the configuration, so a binary format makes sense. There are many ways to have program-writeable text format (netscape/mozilla to name a big GUI app has been doing it for as long as I can remember), so again it's not a user wish but programming laziness the proprietary context allowed > > And they generally (ab)use threading, > > Users want the program to be responsive. Developers want purity of design. Yeah, right, that's why every "liberated" program is described as a pig compared to the FLOSS alternatives. Hint : purity of design translates in perf wins, hacks translate in microbenchmark wins for the first version (and then no one dares updating them for the next ones so all the clever ugly code rots) > > or > > custom plugin, upgrade, and installation systems. > > To reach actual users (not developers) you need more than rpm and yum. Wrong. The percentage of users willing to go the specific update route is vanishingly small, only enthousiasts are willing to spend the time learning an update system per app. Developers will want source tarballs Enthousiasts will want a way to update everything on their system to the latest bling, and are willing to jump through hoops and custom updaters for this. Everyone else (99% of users) just want the stuff to be pre-installed and transparently updated with the rest of the system using the system tools. For all its faults windows update showed pretty conclusively even in the windows world the power of a single unified update service. Regards, -- Nicolas Mailhot -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
