Simon Wilkinson <s...@inf.ed.ac.uk> writes: > On 29 Sep 2010, at 17:25, Jeffrey Altman wrote:
>> 1. Should it be imported into the OpenAFS repository from github as >> an external? >> 2. Should it be required that developers download it and install it >> themselves? > Personally, I have a preference against importing binary build products > into a source repository. This is partly out of neatness concerns, and > partly because of the overhead it places when fetching the > repository. If the products are large, or regularly updated, this can be > a particular problem with git as it can cause the repository to bloat > spectacularly. Yeah, that's my opinion except even stronger. I think separate products should remain separate whenever possible. I'd much rather have people install build dependencies rather than trying to import everything into the OpenAFS repository, regardless of how we go about doing that. With my Debian hat on, I don't have opinions about Windows, but to argue from a similar case, I'm already unhappy with the fact that we had to import bits of Heimdal. It's unavoidable for the kernel build, but it's something we want to do as little of as possible, and the userspace should still use regular Heimdal libraries whenever we can. It makes life for distributions really hard when there are multiple copies of libraries, since security fixes become a huge pain in the ass. > I'd much rather that we provide a website to host build requirements, > and (possibly) a script to populate the tree with any dependencies > before the build commences. I'd don't believe that this is overly > arduous for those who build from source - people building on Unix have > been doing this for years now (we don't ship MIT Kerberos with the Unix > build, for example). Yes. -- Russ Allbery (r...@stanford.edu) <http://www.eyrie.org/~eagle/> _______________________________________________ OpenAFS-devel mailing list OpenAFS-devel@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-devel
