On Tue, May 17, 2005 at 06:20:45AM -0500, Dirk Eddelbuettel wrote: > | No, if you're going to provide a shared library for use by other packages in > | the archive, these are the rules that have to be followed. They are not > | arbitrary requirements; there really is no other way to properly support > | partial upgrades with shared library packages.
> I hear you, but nobody in their right mind would do partial upgrades. Well, one of the effects of supporting partial upgrades is that if an admin has to stop an upgrade in the middle, the system remains usable as long as the packages are in an "installed" state. > | Nor does changing the binary package name mean redundant packages that you > | have to chase after. (It's only the binary package name that should change, > | not the source package name.) The worst effect is that there will be a > Ok, I see. So then it still fills up the users computers as it will create > orphan library packages. Yes, which is also an issue for any other library package if the user is not using tools such as deborphan, debfoster, or aptitude. > | delay while the package goes through NEW processing, and this is minor > | nowadays. > Agreed. > So I guess it would have to become libquantlib-0.3.9 and > libquantlib-0.3.9-dev as with, say libstdc++5 ? Is there a reason to use a versioned name for the -dev package? If not, it's probably fine to leave it as libquantlib0-dev, which may not be accurate, but doesn't seem to hurt anything and is the least disruptive. For most libraries, there's usually no need for versioned -dev packages. -- Steve Langasek postmodern programmer
signature.asc
Description: Digital signature
