Anthony, I can see what your concern is. I hope it can be tempered by some history:
The computer industry has approximately ten years of experience with the development of ABI specifications and test suites. The work done by the ABI+ group, focused on Intel processors, is a direct predecessor of the LSB work, and is itself based on previous ABI efforts. ABI+ and the other efforts have all included both specification and tests. The direct line of descent of both specification and tests starts in 1992, to my knowledge, and possibly even earlier than that. The point is, both the tests and the specification have been developed in conjunction with each other for a long time. The LSB group includes people who were heavily involved in the previous ABI work, so we still have the collective experience we need. The LSB tests are being developed in a tried and true framework by a team of people including some of the technical leaders in previous ABI development. I think most people involved in the LSB work are clear that the specification and tests, plus a sample implementation, must be developed simultaneously and in conjunction with each other. It's an awful lot more work than it appears on the surface but it is proceeding in a well-reasoned fashion. -----Original Message----- From: Anthony Towns [mailto:[EMAIL PROTECTED] Sent: Tuesday, August 21, 2001 3:00 AM To: [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: Re: LSB.fhs failure digest for Debian Sample Implementation On Fri, Aug 17, 2001 at 08:43:55AM +0100, Andrew Josey wrote: > On Aug 17, 10:28am in "Re: LSB.fhs failure ", Anthony Towns wrote: > > It still seems a bit meaningless to have lots of test cases when there's > > not even a prototypical working product (ie, some .lsb packages that > > can be installed on some distro), though. > Without tests we won't know (a) whether the specifications are > useable/correct or (b) whether implementations implement the specifications. > In general tests provide a useful measure of the quality of a > specification, embodying the requirements in code that can be run > and produce measureable results. Without actual lsb packages and an installer, you don't know if the specifications have anything to do with helping ISV's distribute Linux software. With just the SUS and no actual Unix kernel or apps, you wouldn't know if "Unix" was a useful operating system worth being compatible with, or if it was headed off in some completely wrong direction... None of which is to dispute the utility of all these testing things: without them you wouldn't be able to be very confident if people claiming compliance with the LSB were compliant or not, which'd render the LSB completely useless pretty gee darn quickly. But it'd be nice to see some demonstration that the LSB spec is actually useful for what it's meant to be for, rather than just having to take it on faith. I'd've thought it'd be completely trivial to make up a .lsb that'll probably install on future LSB compliant distros, though. Install Red Hat or similar, setup the libs right, build an rpm, and rename it, and that's it. No? Cheers, aj, still stuck on an unimplemented spec being considered "1.0" -- Anthony Towns <[EMAIL PROTECTED]> <http://azure.humbug.org.au/~aj/> I don't speak for anyone save myself. GPG signed mail preferred. ``_Any_ increase in interface difficulty, in exchange for a benefit you do not understand, cannot perceive, or don't care about, is too much.'' -- John S. Novak, III (The Humblest Man on the Net)
