* Stuart Anderson ([EMAIL PROTECTED]) [001024 18:01]: > > Do you (and Nick) mean a C library with programming interfaces?
I'm not. I'm talking about a documented procedural approach, or what I've called a protocol. In other words, "Here's the first step you must take before attempting to install an application. Here's how you search for the dependencies within the file system. Here's the convention you need to use for naming and versioning files so that the search is reliable. Here's how to query the database. Here's the convention for versioning and names for the database so that the search is successful. Here's how to update the database when you cannot rely on the version being reflected by the file name...etc." Please don't hold me to the above, it's just an example off the top of my head to illustrate how I'm thinking. If you want to take the above and build a C API around it, that would be wonderful! But if the above is defined and well thought out, you could recreate the system in any language. The only part of this system that might be tricky is the database. And as I said elsewhere, if the RPM database would serve, then fine - I'm not advocating starting from scratch if we don't have to. I'm simply advocating that we do it right. Finally, let me make one more thing very clear - I'm not limiting my viewpoint to the extremely narrow perspective of what LSB 1.0 is likely to cover. We'd be ostriches if we think that people are only going to build applications that ONLY use those things defined in LSB 1.0. Let's get real and address installation from a more global perspective so we can solve the problem of incompatibilities across versions properly. -Nick -- ********************************************************** Nicholas Petreley Caldera Systems - LinuxWorld/InfoWorld [EMAIL PROTECTED] - http://www.petreley.com - Eph 6:12 ********************************************************** .
