“The Linux Standard Base project aims to keep subtle differences between implementations of the operating system from making applications incompatible across distributions. Last month’s release of LSB 3.2 continues along that road, furthering compatibility and encompassing new standards for multimedia and scripting languages.”
Though it is useful for some thing I don’t see LSB as very useful for many of the things it supposedly standardizes. Its approach seems to be merely to collect every current practice together into one document and declare that they’re all standard. This helps with adoption, since there isn’t much change required, and it may help decrease future divergence, but it strikes me as more of a rubber stamp on whatever people want to do than an attempt to define the way things should be done (you know, like a standard way).
Yes, but if you specified standard config files, aptitude for everyone and GNOME as the standard DE, even if that would benefit Linux as a whole, no distribution would follow your specification. Except maybe your own LSB-approved “standards”-compliant distro.
But what, then, is the point? If a standard doesn’t let you know what you can expect on a system, which in many cases the LSB doesn’t due to its preference for saying “this, or that, or that, or that can be found,” then is it actually useful?
I could describe a standard definition for, say, what a web browser is that describes all capabilities (e.g. supported tags and css behavior) of all current browsers. Would that be better than a narrow definion that none can quite meet (hint: this is what we have now)? If you normalize aberrant behavior by creating a standard that says it’s okay, did you really accomplish anything? As I said originally you might prevent future divergence, but that’s about all.
No, you are not likely to get everyone to agree with and use your standard if you limit what can be considered standard. Is that a problem? Is wide adoption of a set of very broad standards better than small or slow adoption of specific standards? Look at fd.o; they create specs which are often not widely used, but they are very specific and do not merely envelop all current practices.
I beg your pardon?
The LSB is very specific about what to expect on a system. Each library’s symbols are listed, which in case of a C++ library even specifies the name mangling scheme used by the compiler.
Even fd.o specs at draft status are often implemented by at least GNOME and KDE, specs with better completion status pretty certainly are.
I’d say that being used by GNOME and KDE definitely counts as widely used in the area of free software desktops.
EDIT: adding part regarding fd.o specs
Edited 2008-02-28 20:35 UTC