From: Nikodemus S. <tsi...@cc...> - 2003-05-08 13:09:41
|
On Thu, 8 May 2003, Daniel Barlow wrote: > > try to live with org.foobar.yadda.cl-dingdong. If definitely doesn't make > > it easier to see what the package is about. > > It's not supposed to, though. It's supposed to disambiguate packages > - From different people who may want to use similar names (probably > because they weren't aware of the other package already using > whatever name) > > If we lose this, we have to hope that people get smarter about > choosing their package names to avoid conflicts in the first place, > or we have to get stricter about what does or doesn't go into cclan. > Who gets the "md5" package? I think some bit of control is good. I think that packages in this respect can be split into two categories: * Those for which canonical cclan versions make sense. Like md5. I mean, how many alternative md5 packages do you think we need? * Those for which multiple implementations make sense. Like XML parsers. The naming schemes could be: category-canon (eg crypto-md5 <- canonical package) category-canon-name (eg crypto-md5-foobar <- alternative md5 package named foobar) and category-name (eg xml-foobar <- xml tools in called foobar) This is not just a disambiguation issue, IMHO. It's also a classification and distribution issue. If cclan abosulety wants to use a dot-hieracrhy, then I suggest the followwing: cclan.category.name for example: cclan.md5.foobar cclan.xml-parsers.foobar possibly for canonical packages: cclan.ssl Before you ask, there is a reason why I used varying categories for packages: I don't think it's a solved problem. Should md5 be in crypto, security, or is it a category of it's own? Just thinking out loud. Cheers, -- Nikodemus |