It's about time...
Posted Aug 4, 2005 3:41 UTC (Thu) by
jabby (subscriber, #2648)
Parent article:
Our bloat problem
It's about time someone started making more noise about the bloat on the Linux desktop. It's sad when non-profits who can't afford newer computers to run a Microsoft OS more recent than Win98 look to Linux to reuse their older hardware and I have to tell them "Well, theoretically you can, but..."
Even after setting up LTSP on a reasonably powerful machine, it can only power about 4-5 terminals with KDE or GNOME and Firefox before the RAM on the LTSP server is exhausted. And LTSP is less than ideal because sound and local removable storage don't really work, yet (at least I couldn't get them working).
This is a sad commentary on the supposedly inclusive nature of Linux. Even though we all have access to the code, no one is spending time so that we can *all* run it!
Efforts like the RULE Project have sought to create streamlined installs of major distributions, choosing only lightweight apps and desktops. Other efforts include specialized distributions, which thereby isolate the user into a smaller, niche community. Both approaches can be successful, but the applications they use lack the depth of community support of their mainstream (bloated) cousins.
Nevertheless, these apps prove that the bloat is not necessary. Anyone who has used the ROX-Filer file manager knows that it is possible to write a GUI filemanager that runs blazingly fast on an old 586 box and uses very little of that precious RAM.
I agree with Jon's observation about the temporary rush for features to the detriment of performance. It is the bloated toolkits and all of the supporting libraries for the major desktops that have both allowed a sudden proliferation of apps and features as well as saddled those apps with unrealistic hardware requirements. My key example is KWrite. It's a simple text editor with a few extra features. It takes *seconds* to load on my 1GHz Athlon machine with 256MB of RAM. I can't imagine trying to use KDE 3.4 on a 586 with 32MB...
I like the suggestion in the first thread of creating a "bloat chart" so developers can measure and remedy their bloat. But I think developers and distribution maintainers really need to look at dependencies a lot closer and try to remove as many as possible. Rather than linking to a library because it's there and you can, really evaluate the trade-off between the functionality you're getting and the bloat you're dragging along to get it. If jg is right and most of those libraries don't need to be loaded, then breaking those false dependencies would go a long way toward solving the immediate logjam.
(
Log in to post comments)