On Thu, Nov 05, 2009 at 10:55:25AM +0000, Iain Lane wrote: > I'm forwarding your mail to the Debian Haskell Group and to kaol, Would you believe that I'm subscribed. Personally, I think that moving to support shared libraries in time for the next Debian release is not worth it. From the top of my head, a few things: AFAIK when GHC's upstream says Linux, they mean Linux on i386 and amd64. That's not good enough for us. Even if there wasn't that, it would mean that we'd need to change the Haskell library packaging scheme for all packages. We'd need to introduce libghc6-$(NAME)$(SONAME) packages along with the -dev and -prof packages. That'd mean pulling everything through NEW, as well as thinking of what that $(SONAME) would actually be like. FTP masters have already grumbled about the amount of binary packages that Haskell libraries generate. There's always the option of merging -dev and -prof packages, which could IMHO be defended since they're only useful on developers' machines and supposedly they have the disk space to devote to that. Keeping ghc6-prof around is still worth it since anyone can feel that 141MB. Or perhaps we could generate profiling info, include it in libraries' packages and keep it compressed and only uncompress it in ghc6-prof's trigger. However, I wouldn't go into any of that for 6.12.1 due to the impending release. I'm checking through 6.12 rc1 currently and I'm counting on getting it ready by November-December when 6.12.1 is due to be released and get it packaged soon after that. But I'm not comfortable with revamping the Haskell library policy this much in this time frame. I'd rather leave it for Squeeze+1 and let the shared library support mature enough during that time. We'll likely be out of sync with Haskell Platform in Squeeze. I want GHC 6.12.1 in it and the freeze is likely to start before the plaftorm will include that. I'm not overly concerned about that.
Description: Digital signature